` Universita degli Studi di Macerata Dipartimento di Istituzioni Economiche e Finanziarie

Do M&As in the EU banking industry lead to an increase in performance?

Elena Beccalli, Pascal Frantz

Quaderno di Dipartimento n. 50

Dicembre 2008

Do M&As in the EU banking industry lead to an increase in performance? Elena Beccalli, Pascal Frantz

Abstract
This paper investigates whether M&A operations inuence the performance of banks. Using a sample of 714 deals involving EU acquirers and targets located throughout the world over the period 1991-2005, we investigate whether M&A operations are associated with improved performance (measured using both standard accounting ratios and cost and alternative prot X-eciency measures). Despite the extensive and ongoing consolidation process in the banking industry, we nd that M&A operations are associated with a slight deterioration in return on equity, cash ow return and prot eciency and with a marked improvement in cost eciency. Hence, the improvements in cost eciency appear to be transferred to bank clients. These changes in performance are directly attributable to the M&A operations, and would not have occurred in their absence. Moreover, these changes exhibit a particularly negative trend for cross-border deals to testify the importance of geographical relatedness in order to achieve better post-M&A performance. The environmental and bank-characteristics that make a deal successful or unsuccessful are nally identied.

Elena Beccalli, Universit` degli Studi di Macerata and London School a of Economics.

E-mail: e.beccalli@lse.ac.uk. Pascal Frantz, London School of Economics. E-mail: p.frantz@lse.ac.uk.

Introduction

This paper is part of a research project - promoted by Arel (Agenzia di ricerche e legislazione founded by Nino Andreatta) and sponsored by UniCredit - co-ordinated by Paolo Gualtieri. The author wishes to acknowledge the constructive comments oered by the discussant and participants at the Conference on Mergers and Acquisitions of Financial Institutions Federal Reserve Bank, FDIC (30th November 2007, Washington). The authors are grateful for the research assistance oered by Francesco Pisano and Livia Spata, and for the comments oered by Philip Molyneux, Allen Berger, Giovanni Petrella, Agostino Fusconi and Francesco Cesarini. This paper investigates the eect of mergers and acquisitions on the performance of banks and explores the sources of any merger-induced changes in performance. It is motivated by the relative dearth of empirical evidence on the impact of mergers and acquisitions involving European banks. Overall the handful of studies on merger and acquisition (M&A) activities in the EU banking industry provides mixed results. For instance, Altunbas and Ibanez (2004) report that bank mergers taking place in the EU banking industry between 1992 and 2001 do lead on average to improved accounting protability. Altunbas, Molyneux, and Thornton (1997) provide empirical evidence suggestive of limited opportunities for cost savings from large mergers in the banking industry. Vander Vennet (2002) reports a limited improvement in prot eciency but not in cost eciency with reference to cross-border deals only. This inconclusive evidence appears counterintuitive given that an intensive process of M&A operations transformed the banking industry in the US over the last decades (DeLong and DeYoung, 2007), and that the pursuit of a further integration trough cross-border M&A operations in retail banking is one of the main objectives pursued by the European Central Bank in the EU (Trichet, 2007). The main aim of our paper is to use a comprehensive approach, involving cost eciency, prot eciency, and accounting ratios, in order to test directly whether M&As involving European banks did lead to improvements in performance between 1991 and 2005. To our knowledge, this is the rst study involving a large sample of EU acquiring banks in deals with target banks located throughout the world (including, among the others, US and EU banks). None of the previous studies compare the evidence from all the performance measures (accounting ratios, cost eciency and prot eciency). None of the existing studies disentangle the total change in performance into the part due to the M&A operation itself and the part that would have occurred anyway. Our paper therefore aims to investigate the impact of M&A operations on accounting 1

protability measures and on (cost and alternative prot) X-eciency for a large sample of 714 deals with EU acquirers and targets located in any country of the world over the period 1991-2005. In particular, it extends and integrates the existing literature by enlarging the geographical coverage of the sample, by contemporaneously testing several performance measures, and by distinguishing the part of the change in performance due to the M&A itself. In spite of the extensive and ongoing consolidation process in the banking industry, we nd that M&A operations are associated with a slight deterioration in return on equity, cash-ow returns, and prot eciency and a pronounced improvement in cost eciency in a period of 5 to 6 years following the deals. Hence, the improvements in cost eciency appear to be transferred to bank clients rather than to bank shareholders. Interestingly, these changes in performance are directly attributable to the M&A operations and would not have occurred in the absence of any M&A operation. Moreover, these changes exhibit a particularly negative trend for cross-border deals: in domestic deals, cost eciency improves more markedly than in cross-border deals whilst returns on equity and prot eciency remain unchanged instead of diminishing. This highlights the importance of geographical relatedness in order to achieve better post-M&A performance. Finally, in the years before the M&A operation, target banks exhibit weaker performance than acquirers in terms of prot eciency, cash-ow returns, returns on equity, personnel expenses and operating costs. Besides, banks involved in M&A operations (both acquirers and targets) are more ecient and protable than their peers not involved in M&A operations. Furthermore, an important set of institutional, regulatory, bank-specic and deal-specic variables has a signicant inuence on changes in cost and prot eciency. The management of acquiring banks should tend to direct investments to those countries that guarantee better regulatory quality together with higher freedom from government. Moreover, to achieve positive changes in eciency in the medium-term, transactions should be paid in equity (not in cash) and result in a combined bank with a higher focus on traditional banking activities. Finally, a number of bank and environmental characteristics make a deal successful or unsuccessful. A higher likelihood of an unsuccessful M&A (in terms of prot eciency) is associated with larger size and equity capital of the acquirer, a higher risk (of both the acquirer and the target), a larger diversication of the target, a higher freedom from government and regulatory quality in the country of the acquirer, and a lower regulatory quality in the country of the target. A higher likelihood of an unsuccessful M&A (measured in terms of cost eciency) is associated with smaller size of the acquirer, a 2

smaller equity capital of the acquirer, and a higher loan loss provision (of both the acquirer and the target). The paper is organised as follows. Section 2 provides a literature review and notes the motivation for our study. Section 3 outlines the methodological approach, and illustrates the sample and data. Finally section 4 describes the empirical results, and section 5 concludes.

Literature and motivations

Surprisingly, the available empirical evidence suggests that M&A operations in the US banking industry have not had a positive inuence on performance (DeLong and Deyoung, 2007; Amel, et al., 2004; Berger, Demsetz, and Strahan, 1999). Overall these studies provide mixed evidence and many fail to show a clear relationship between M&As and performance. Some of the previous literature has examined the impact of M&A operation on cost eciency as measured by simple accounting cost ratios (Rhoades, 1990, 1993; Pillo, 1996; DeLong and DeYoung, 2007), the impact on cost X-eciency (Berger and Humphrey, 1992; DeYoung, 1997; Peristiani, 1997; Berger, 1998; Rhoades, 1998), the impact on protability ratios such as ROE and ROA (Berger and Humphrey, 1992; Pillo, 1996; Knapp et al., 2006; DeLong and DeYoung, 2007), and the impact on prot X-eciency (Akhavein et al., 1997; Berger, 1998). Neither the earlier studies nor more recent analysis nd evidence of clear positive eects of M&A operations on the performance of US banks. Most of the empirical evidence on the impact of M&A operation on Xeciency relates to the US banking sector and to the estimation of cost eciency only. The evidence shows that very minor or absent improvements in cost X-eciency were achieved by M&A operations during the 80s (De Young, 1997; Peristiani, 1997). By using a thick frontier approach on a sample of 348 deals, DeYoung (1997) nds that 58% of the banks in the sample generated cost eciency. Interestingly, mergers in which the acquiring bank had recent experience with acquisitions were more likely to generate post-merger cost eciency gains. As regard to 4,900 transactions occurred between 1980 and 1990, Peristiani (1997) suggests that acquirers failed to improve X-eciency after the merger, but acquiring banks experienced moderate gains in scale eciency relative to a control sample. As regard to the 90s, there is mixed empirical evidence (Rhoades, 1998; Berger, 1998). For nine deals involving relatively large banks during the early 1990s, Rhoades (1998) nds that four of the nine mergers were clearly successful in improving cost X-eciency but ve were not, although all nine of the mergers resulted 3

in signicant cost cutting. For deals involving both large and small banks from 1990 to 1995, Berger (1998) instead nds very small improvements in cost X-eciency. Although most of the studies focus on cost eciency, few attempts have been done to estimate the eects on prot eciency for US banks (Akhavein et al., 1997; Berger, 1998). By investigating US megamergers (i.e. both partners with more than $1 billion in assets) over the period 1980-1990, Akhavein et al. (1997) nd improvements in prot eciency (+16% in comparison with other big banks). Most of the improvement comes from a better risk diversication and increased revenues, including a change in the output composition from securities in the bank portfolio to loans. The highest improvement is recorded for the banks with the lowest eciencies prior to the merger, which therefore had the greatest capacity for improvements. Berger (1998) nds similar results in a study that includes all US bank mergers, both large and small, from 1990 to 1995. The handful of studies on the M&A activities in the EU banking industry also seem to conclude that performance improvements are seldom realised. These studies have examined the impact of M&A operation on cost X-eciency (Vander Vennet, 1996, 2002; Altunbas, Molyneux and Thornton, 1997), the impact on protability ratios such as ROE and ROA (Vander Vennet, 1996; Altunbas and Ibnez, 2004), and the impact on prot Xa eciency (Huizinga et al., 2001, Vander Vennet 2002). Altunbas, Molyneux and Thornton (1997) estimate a hybrid translog cost function for a pooled sample of French, German, Italian and Spanish banks for 1988 only. Their results suggest only limited opportunities for cost savings from big-bank mergers, and instead an increase in total costs appears more likely. With regard to a sample of 492 M&A operations related to EU banks over the period 1988-1993, Vander Vennet (1996) shows that domestic mergers among equal-sized partners signicantly increase the accounting protability of the merged banks, whereas improvements in cost eciency are observed only for cross-border acquisitions (and not for domestic operations). Domestic takeovers are found to be inuenced predominantly by defensive and managerial motives such as size maximization. For a small sample of 52 bank mergers over the period 1992-1998, Huizinga et al. (2001) nd that the cost eciency of merging banks is positively aected by the deal, while the relative degree of prot eciency improves only marginally. In a specic focus on cross-border deals among EU banks, Vander Vennet (2002) refers to a sample of 62 operations executed by banks headquartered in the EU, Norway and Switzerland between 1990 and 2001. In the short period after the deal, he nds a limited improvement in prot eciency, but no improvement in cost eciency. His analysis also reveals large dierences in the cost and prot e4

ciency of the acquirer and target pre-deal. Altunbas and Ibnez (2004) with a regard to 262 deals taking place in the EU banking sector between 1992 and 2001 nd that, on average, bank mergers resulted in improved accounting protability (ROE). Several explanations for this puzzling evidence have been provided: absence of best-practices guidelines for planning and executing increasingly large and complex acquisitions (DeLong and DeYoung, 2007), failure in considering the mean-reversion behaviour in industry-adjusted performance (Knapp et al., 2006); longer time (up to ve years) needed to realise eciency gains, mergers and acquisitions leading to more favourable prices for consumers (Focarelli and Panetta, 2003), diculties in integrating broadly dissimilar institutions (Altunbas and Ibnez 2004; Vander Vennet, 2002), a increased costs associated with changes in post-merger risk proles and business strategies (Demsetz and Strahan, 1997; Hughes et al., 1999). Nevertheless all the above studies just refer to the overall change in performance by comparing in a dynamic analysis (according to the denition by Berger, 1998 and 1999) the post-M&A performance with the pre-M&A performance. However, some of this dierence could be due to a continuation of rm-specic performance before the merger or to economy wide and industry factors, as stated by Healy et al. (1992). Healy et al. (1992) however do not specically investigate the banking industry and just refer to the impact on operating cash ow returns of the 50 largest US mergers over the period 1979 and 1984. In short, none of the above studies consider a large sample of EU acquiring banks involved in deals with target banks located throughout the world; none compare the evidence from all the performance measures; and none disentangle the total change in performance into the part due to the M&A operation itself and the part that would have occurred anyway. Our paper therefore aims to extend and integrate the existing literature by enlarging the geographical coverage of the sample, by contemporaneously testing several performance measures, and by distinguishing the part of the change in performance due to the M&A itself. These elements constitute the main novelties of this analysis.

Methodology

Our study uses a variety of ways to investigate the relationship between bank performance measure in the pre- and post- deal period. The initial approach to test this relationship follows the traditional banking literature on M&A and performance measures (reviewed above). By conducting ANOVA tests, 5

we thus compare: 1. Performance measures for target and acquirer in the pre-M&A period; 2. Performance measures for banks involved in M&A operations and banks not involved in any M&A operations.; 3. Performance values post-merger for combined banks resulting from the M&A deal and weighted averages of the performance measures of the targets and acquirers prior to the M&A deals (with total assets as weights). In order to take into account the fact that any performance measure can be aected by both bank-specic inuences and industry-wide trends, we introduce industry-adjusted measures. This industry-adjusted performance measure, also referred to as abnormal performance, is derived as the dierence between the performance measure for each M&A bank minus the (average) performance of the industry control sample (all other banks never involved in any M&A operations, matched on the basis of the country of the M&A bank and the year under investigation). Performance measures used in this paper refer either to accounting profitability measures (estimated by annual ROE and cash ow returns) or to global measures of operational eciency (estimated by both cost and alternative prot X-eciency). The statistical signicance of the industry-adjusted gures is based on t-statistics, and on the non-parametric Wilcoxon test to assess the signicance in the case of non-normality. To ensure that industryadjusted gures are not driven by outliers, the portion of positive cases is also reported. The dynamic analysis covers a medium-long term period either starting six years before and ending six years after a deal (6B-6A) or starting three years before and ending three years after (3B,3A). For each of the years surrounding the deal, we calculate the mean value of the relevant ratios for the banks involved. For accounting ratios we also calculate median values, as they are more susceptible to outliers. The year of the deal itself is left out of the analysis as it can be considered as a transition period strongly aected the accounting practices regarding M&As. The measure of the change in performance as described here above provides some informative (but not conclusive) evidence about the impact of M&A operations on performance. The dierence in the performance priorand after- the deal, could also be due to economy-wide and industry factors, or the continuation of rm-specic performance before the operation (Healy et al., 1992). Accounting measures typically move to the industry mean in a process known as mean reversion (Fama and French, 2000; Knapp et al., 6

2006). To further investigate the relationship between pre- and post- deal industry-adjusted performance, we hence split the overall change into its several determinants by using the following cross-sectional regression: AdjP erM &A,post = + AdjP erM &A,pre + (1)

where AdjPer is the average annual industry-adjusted performance for each M&A (as previously noted, performance measures are both accounting values and X-eciency estimation). AdjPer M &A,post refers to the post-M&A period (i.e. to each of the 6 years after the deal), whereas AdjPer M &A,pre refers to pre-M&A period, known as base period, which represents the weighted average of the performance measure of the target and acquirer in the 3 years (or alternatively in the 6 years) prior to the M&A. Following the interpretation of Healy et al. (1992), the slope coecient captures any correlation in performance between the pre- and post- M&A years so that AdjP erpre,M &A measures the eect of the pre-M&A performance on the post-M&A performance. The intercept is therefore independent of pre-M&A performance and hence measures the impact of the M&A operation on performance. Note that in order to investigate the size eect on the variation in performance determined by the M&A, we use the following OLS regressions: AdjP erM &A,post = + AdjP erM &A,pre + SizeA,pre + SizeT,pre + (2) Furthermore, to control for the determinants of the change in performance, several regulatory, bank-specic and deal-specic variables are used as control variables. The estimated regression equation is: AdjP erM &A, pre vs post = + (CVA, pre , CVT, pre , CVC, post ) + where CV are the control variables: 1. deal-specic: dummy for the year of the deal announcement (either before or after year 2000), dummy for cross-country and domestic deals, dummy for the method of payment (cash vs. equity); 2. bank-specic: size (where size is measured as the natural logarithm of total assets) of acquirer (A), target (T), and combined bank resulting from the deal (C); risk of the business (where risk is measured by the standard deviation of ROE) of acquirer, target, and combined bank; (3)

proportion of traditional banking (measured by loans/total assets) for acquirer, target and combined entity1 ; 3. regulatory and institutional (of the countries of the acquirer and target): freedom from government (dened to include all government transfers and state-owned enterprises), regulatory quality (the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development), concentration index of the banking industry (total value of assets of the ve biggest banks). Finally, to compare deals that work and those that do not work, we divide deals into successful and unsuccessful deals. Successful (unsuccessful) deals are deals in which the change in eciency (measured as adjusted eciency after the deal minus adjusted eciency before the deal) is comprised in the rst (fth) quintile. The dierentiation of successful and unsuccessful deals is based on the following characteristics of the acquirer and target: a) size (proxied by natural logarithm of total assets); b) capital (proxied by the normalised value of equity); c) risk (proxied by standard deviation of ROE); d) diversication (proxied by o-balance sheet / total assets); e) quality of loan portfolio (proxied by the normalised value of loan loss provision); f) freedom from government; g) regulatory quality; h) year of the deal announcement (before vs. after 2000).

3.1

Data set and sample

The data set is obtained by combining three sources: Thomson One Banker M&A for data on the M&A operations; Thomson Financial Datastream for prices of listed banks, benchmark, and economic indexes; Bankscope for balance sheet and prot and loss data of the banks involved in M&A operations (M&A sample) and of banks not involved (control sample). The sample is limited to credit institutions as dened in the Second Banking Directive (excluded are deals involving securities rms, insurance companies, investment banks or nance companies). It comprises M&A deals announced between 1/1/1991 and 31/12/2005 in which the acquirer is a EU bank and the target is a bank operating in any country of the world. The initial M&A sample refers to 970 observations, but the nal one contains 714
We also tested for other variables on bank-level characteristics: quality of loan portfolio (proxied either by total problematic loans or by loan loss provision) and equity capital. However these bank-level variables are not proven to be statistical signicant either for cost or for prot eciency changes.
1

deals (394 domestic and 320 cross-border transactions) for which full nancial information about the participating banks is available. It is a unique sample, bigger than any other sample used for the analysis of M&A operations in the EU banking industry. Table 1 shows the total number of deals constituting the sample in each country and year, and the total panel under observation. Table 2 highlights the home country of target and acquirer in cross-border deals over the years under observation. In any given year, the control sample consists of all banks never engaged in any merger or acquisition over the life span of this study - that match the nationalities of the acquirers and targets during a specic year. As shown in Table 3, the control sample consists of 7,963 observations over the period 1991-2005. For any M&A deal, there is a control for both the acquirer and the target. By default, in the X eciency studies, the control for any performance measure related to an acquirer (target) is the mean performance of all the banks in the same country than the acquirer (target), and same year. Accounting ratios however do exhibit signicant skewness. In accounting studies, by default, the control for any performance measure related to an acquirer (target) is hence the median performance of all the banks in the same sector of activity than the acquirer (target), in the same country, and same year.

3.2

Accounting ratios as performance measure

This paper introduces two main accounting-based ratios in order to assess performance as far as shareholders are concerned: return on equity (ROE) and cash-ow return (CFR). A rms ROE is dened by default as the ratio of net income over the book value of equity obtaining at the beginning of the nancial year. This ratio relies on the properties of accrual accounting in order to assess performance. Whilst widely used, this ratio is however aected by the method of accounting for the acquisition or merger, by the method used to nance the M&A (cash vs shares) and by other discretionary accounting choices made by managers. Hence we also assess performance through cash-ow returns. A rms cash-ow return is dened by default as the ratio of operating cash-ow over the market value of equity obtaining at the beginning of the nancial year. The operating cash-ow is furthermore derived as net revenue (interest revenue, commission income, and trading income) less cost of generating revenues (interest expense, commission expense, and trading expense), less personnel expenses, and other administrative expenses. The cash-ow return performance measure, unlike return on equity, is unaected by depreciation and goodwill. This market-based performance measure is however aected by changes in expectations about future cash9

ows. Because the market values used in the cash ow returns are aected by the M&A, we subtract the market premium from the market value of equity in order to measure performance following the M&A deal. Regardless of the performance measure used, we do exclude the year in which the acquisition or merger is taking place because of dierences between the acquisition and merger methods in timing the consolidation of the acquirer with the target.

3.3

Operating eciency as performance measure

In addition to traditional accounting ratios, we introduced a more advanced measure of operational productivity at the global level, the so-called Xeciency (Leibenstein, 1966). It is generally accepted in the empirical banking literature that frontier analysis provides an overall, objectively determined, numerical eciency value (known as X-eciency) and ranking of rms that is not otherwise available (Berger and Humphrey, 1997). This attribute makes frontier analysis particularly valuable in assessing and informing government policy regarding nancial institutions, particularly in the context of mergers and acquisitions. X-ineciency is a measure of managerial best practice, and represents the distance of the position of equilibrium of each bank from the optimal operative frontier. X-eciency can be framed as: 1. Cost eciency, which provides a measure of how close a bank is to the cost sustained by the best practice bank to produce a given mix of outputs (assuming that the banks are operating under the same conditions). A bank is said to be cost minimising when it consumes a lower quantity of inputs for the production of a given amount of outputs or, in other words, produces the same amount of outputs using less inputs and, in this way, enjoys a cost advantage; 2. Prot eciency, which provides a measure of how close a bank is to the realisation of the maximum level of prot given its level of outputs (generally known as alternative prot X-ineciencies). A bank is said to be prot maximising when it produces a greater quantity of outputs given the amount of inputs employed. It indicates that the bank produces more outputs (or outputs of a higher quality) using the same amount of inputs and, thus, is able to apply a price premium. Following Berger and Mester (1997), we prefer to choose a parametric approach as opposed to a non-parametric approach as it is particularly eective in representing the concepts of cost and prot eciency. We employ the standard Stochastic Frontier Approach (SFA) to generate estimates of 10

cost and alternative prot eciencies for each bank over the years 1991-2005 along the lines rst suggested by Aigner et al. (1977). Specically, we employ the Battese and Coelli (1995) model of a stochastic frontier function for panel data with rm eects which are assumed to be distributed as truncated normal random variables (=0) and are also permitted to vary systematically with time (see for more details on the SFA methodology Coelli et al., 1998). The functional form for the frontier is a Fourier exible (FF) form, which is a global approximation that dominates the conventional translog form. The characteristic of global approximation is particularly important in the case of the study of the eects of M&As on banks around the world, because the scale of banks, the diversication of their products and services and the levels of their ineciency are often heterogeneous (see, for example, Gallant 1981; McAllister and McManus 1993; Mitchell and Onvural, 1996). It combines the stability of the translog specication around the average of the sample and the exibility of the Fourier specication for the observations that are far from the average. The FF functional form, including a standard translog and all rst- and second-order trigonometric terms, as well as a two-component error structure is estimated using a maximum likelihood procedure. This is specied as follows: ln T C = 0 + +1 2 +
3 3 i=1 3 3

i ln Qi +

3 j=1

j ln Pj + 1 T + 1 ln E+
3 3

ij ln Qi ln Qj +
3 j=1

ij ln Pi ln Pj + 11 ln E ln E
3 i=1

i=1 j=1 3

i=1 j=1

ij ln Qi ln Pj + [ai i=1 3 3
i=1 j=1 3

j1 ln Pj ln E +

i1 ln Qi ln E +

i=1 j=1

+ +

cos (zi ) + bi sin (zi )] + [aij cos (zi + zj ) + bij sin (zi + zj )] +

(4) where: TC is a measure of the total cost of production (including labour costs, depreciation, other operating and administrative costs and interests paid on deposits); Qi represent bank outputs (with 1.0 added to avoid taking the log of zero): Q1 = total loans, Q2 = securities, Q3 = o balance sheet business; Pi are bank input prices for labour (= personnel expenses/total assets), price for loanable funds (= interest expenses/total deposits) and price for physical capital (depreciation and other capital expenses/xed assets). 11

Equity capital (E) is included to control for dierences in bank risk preferences (Mester, 1996). zi are the adjusted values of the log output lnQi such that they span the interval [0.1. 2., 0.9. 2.] to reduce approximation problems near the endpoints.2 is the two-component stochastic error term. , , , , , , , , , are parameters to be estimated. While there continues to be debate about the denition of input and output used in the function, we follow the traditional intermediation approach of Sealey and Lindley (1977), in which inputs (labour, physical capital and deposits) are used to produce earning assets. Two of our outputs (loans and securities) are earnings assets, and we also include o balance sheet items as a third output.3 The alternative prot function has the same specication as the above, the only dierence being that the dependent variable is replaced with ln prots ( + ), as specied in Berger and Mester (1997). is a measure of operating prots (interest revenues + commission income + trading income total costs). To exclude negative values, + = + min + 1 , where min is the absolute value of the minimum value of prots in the sample. Being aware of the problems associated with international comparisons of banking eciency (Berger, 2007)4 , in this paper, we adopt a common cross-country frontier for banking industries across the world with the inclusion of environmental variables (along the lines of Dietsch and Lozano-Vivas, 2000; Kwan, 2003). This choice to use a common frontier with environmenzi = i (ln Qi + wi ), where i and wi are scaling factors, limiting the periodic sine and cosine trigonometric functions within one period length 2 (see for a discussion: Gallant, 1981; for an application: Mitchell and Onvural, 1996). 3 Although o balance sheet items are not earning assets, they do represent an increasing source of income for all types of banks and are therefore included in order to avoid understating total output (Jagtiani and Khanthavit, 1996). 4 Berger (2007) crystallizes the three types of cross-country eciency comparisons used in the banking literature: 1) based on nation-specic frontiers, 2) based on a common frontier, 3) based on foreign-owed versus domestically owned banks in the same nation using a nation-specic frontier. Berger (2007) highlights the following. The rst category nation-specic frontiers - cannot be used to draw any conclusion as to whether banks in one country are more ecient than banks in other countries because they are measured against dierent frontiers. The second category common frontier (Dietsch and LozanoVivas, 2000; Kwan, 2003) - may be problematic because it may be virtually impossible to control for all the environmental variables of dierent countries. Moreover, even if all of the environmental dierences do not exist or are well controlled for, the eciency of institutions within their own country can dier from their eciency as foreign entities in other countries. The third category (Berger et al., 2000) generally resolves the main issue of the rst two categories by analysing separately foreign and domestic banks competing in the economic environment of the same country. Nevertheless even this category is not free of measurement diculties (i.e. sample selection issues and impossibility to measure the cross-subsidies received by foreign banks from home-nation banks).
2

12

tal variables draws from the empirical evidence provided in Beccalli (2004), which suggests that the common frontier with environmental variables delivers consistent X-eciency estimates with the Berger et al. (2000) method5 . Real growth in GDP is used here as a country-specic control variable in the panel6 . This model controls for environmental dierences across countries and investigates the eects of these variables on measured eciency. This methodology essentially allows for a rm-specic and time-varying intercept shift in the distribution of the ineciency term, and this intercept shift is itself a function of the exogenous environmental variables that vary across countries (Battese and Coelli 1995). This study applies Fourier terms (both for the cost frontier and the alternative prot frontier) only for the outputs, leaving the input price eects to be dened entirely be the translog terms (see Berger, Leusner, and Mingo, 1997; Mitchell and Onvural, 1996; Gallant, 1982). Moreover, the usual input price homogeneity restrictions are imposed on logarithmic price terms only, and not on the trigonometric terms (as in Altunbas, Gardener, Molyneux, Moore, 2001). Accordingly, TC, P1 and P2 are normalised by the price of physical capital, P3 . Finally, all the values are expressed in real terms (GDP deator for each country with 1991 as a base year).

Empirical results

We rst examine unadjusted performance (cost eciency, prot eciency, accounting protability and their determinants) for acquirers and targets in each of the six years before the deal. The values highlighted in Table 4
Beccalli (2004) compares and reconciles the three methods for cross-country comparisons of the cost eciency with reference to the UK and Italian investment rms. In particular, Beccalli (2004) both controls for environmental dierences across countries by integrating environmental variables into the denition of the common frontier, and compares the X-eciency of foreign versus domestic institutions operating within the borders of dierent home countries to examine the extent to which nancial institutions are able to monitor and control their subsidiaries operating in other nations. The two methods produce consistent results. In contrast, the traditional method for eciency comparison based on nation-specic frontiers produces dierent results. 6 We have also tested for a larger set of control variables (real growth in GDP, population and concentration ratio). Between these frontiers (including either one control variable or the set of control variables), we found large, positive, and signicant rank order correlation. This correlation suggests that the eciency ordering generated by the common frontier with one environmental variable is consistent with the eciency ordering generated by the common frontier with several environmental variables. Thus we have chosen to minimise the number of control variables and have kept the specication with a country dummy only.
5

13

show that the level of prot eciency is higher for acquirers in comparison with targets in each of the 6 years before the deal (and the dierence is statistically signicant): the higher values for the acquirers range between 1.3% (one year before the deal) and 3.2% (six years before the deal). The level of cost eciency tends to be higher for acquirers than for targets in most of the years before the deal but not statistically signicant. Interestingly, instead, the determinants of cost eciency (labour costs and operating costs) show a clearly better performance for acquirers in comparison with targets: these costs are always lower for the acquirer in comparison to the target. In particular, personnel costs of the acquiring banks are on average 3.7% to 4.7% lower than the personnel costs of the acquired banks. In the remaining part of this section, we will control for the performance of acquirers and targets peers. (Note that we will present the evidence on the accounting measures rst and then move to the results on eciency). To investigate performance as far as shareholders are concerned, we use the return on equity (ROE), where equity is measured at the beginning of the nancial year. As shown in Table 5 (Panel A), acquirers do outperform their peers in each of the ve years prior to the mergers and acquisitions by 2 to 3%. There is also some evidence reported in Table 5 (Panel B) suggesting that targets do outperform their peers in the two years prior to the mergers and acquisitions7 . Acquirers furthermore outperform targets in a period starting ve years prior to the mergers and ending three years prior to the mergers [Table 5 (Panel C)]. As shown in Table 5 (Panel D), there is not much evidence that rms engaging in M&A do outperform their peer post-merger (rst year only). There is furthermore evidence suggesting that rms engaging in mergers and acquisitions experience a decrease in their performance post-merger. As shown in Table 5 (Panel E), in the ve years following the mergers, the median industry-adjusted ROE is about 1%. This compares with a median weighted average of the acquirers industry-adjusted ROE and targets industry-adjusted ROE of about 2% in the ve years prior to the merger. The studys main ndings so far, superior bottom-line performance by acquirers pre-merger and lack of evidence of any increase in bottom-line performance post-merger, are robust to alternative specications of return on equity and peer performance. For example, these ndings still obtain if return on equity is derived on an average basis (that is, if equity is measured as the average of the beginning of the nancial year and end of the nancial year values) or if peer performance is derived as the average (as opposed to the median) return on equity of all banks in the same year, sector of activity,
7

The latter result is however not robust to alternative specications of ROE.

14

and country8 . Any decrease in post-merger industry-adjusted ROE may however not be due to the merger or acquisition. In order to control for the eect of pre-merger performance on post-merger performance, we regress post-merger industry-adjusted ROE on pre-merger industry-adjusted ROE, the regression intercept capturing the direct eect of the merger on performance. As shown in Table 6, the regression intercept is negative and statistically signicant in the second, third, and fourth year following the merger. The empirical evidence hence suggests that the M&A operation has a negative impact on returns on equity. We then distinguish between domestic and cross-border mergers and acquisitions. Acquirers are shown to outperform targets prior to domestic mergers and acquisitions [Table 7 (Panel A)]. In contrast, there is no evidence suggesting that acquirers outperform targets prior to cross-border mergers and acquisitions [Table 7 (Panel A)]. Banks engaging in cross-borders mergers and acquisitions however experience a decrease in their performance from the second to the fth year following the mergers and acquisitions [Table 7 (Panel B)]. In contrast, there is no statistically signicant evidence suggesting that banks engaging in domestic mergers and acquisitions experience a decrease in their performance post-merger following the mergers and acquisitions [Table 7 (Panel B)]. The superior returns on equity experienced by acquirers pre-merger are driven by superior net margins as opposed to superior asset turnover [Table 8 (Panel A)]. In contrast, compared with their peers, targets suer from lower asset turnover in each of the four years prior to the mergers and lower net margins in the two years prior to the mergers [Table 8 (Panel B)]. There is no evidence of any statistically signicant improvement in industry-adjusted asset turnover or net margin post-merger [Table 8 (Panel C)]. Compared with their peers, acquirers have a lower personnel expense as a function of revenue in each of the ve years prior to the mergers and acquisitions [Table 9 (Panel A)]. This is also true for targets in some of the earlier years prior to their acquisitions [Table 9 (Panel B)]. The ratio of personnel expense over revenue is however increasing post-merger [Table 9 (Panel C)]. The same picture arises when analysing the ratio of other administrative expenses over revenue. In order to control for the eect of pre-merger performance on post-merger performance, we again regress post-merger industry-adjusted performance on pre-merger industry-adjusted performance, the regression intercept capturing the direct eect of the merger on performance. In regressions involving
8

Empirical evidence on robustness is available from the authors on request.

15

either net margin or asset turnover, the regression intercepts are negative and statistically signicant regardless of the length of the post-merger interval selected [Table 10 (Panels A and B)]. In regressions involving the ratio of personnel expense over revenue, the regression intercept is negative and statistically signicant when post-merger performance is measured over a period of three years selected [Table 10 (Panel C)]. In regressions involving the ratio of other administrative expenses over revenue, the regression intercept is positive and statistically signicant when post-merger performance is measured over a period of two years but becomes negative and statistically signicant when post-merger performance is measured over a period of ve years [Table 10 (Panel D)]. The empirical evidence hence suggests that the M&A operation has a negative impact on asset turnover and net margins but may have a positive eect on personnel and other administrative expenses. We then turn our attention to cash-ow returns. As shown in Table 11 (Panel A), acquirers do outperform their peers in each of the ve years prior to the mergers and acquisitions. There is however no evidence suggesting that targets do outperform their peers in any of the ve years prior to the mergers and acquisitions [Table 11 (Panel B)]. Acquirers furthermore outperform targets in the two years prior to the mergers and acquisitions [Table 11 (Panel C)]. There is furthermore strong evidence suggesting that rms engaging in mergers and acquisitions experience a decrease in their performance post-merger [Table 11 (Panel D)]. In order to control for the eect of pre-merger performance on post-merger performance, we regress post-merger cash-ow returns on pre-merger cashow returns. Regression intercepts are negative and statistically signicant over many post-merger periods, especially when the market value of equity is measured as the average of the beginning and end of year values [Table 10 (Panels E and F)]. The empirical evidence hence suggests that that the M&A operation has a negative impact on cash-ow returns. In all the empirical tests so far, the performance of any bank involved in an M&A has been adjusted for country, year and sector of activity. In order to make sure that our results are not driven by failing to adjust for size, we investigate the correlation between accounting performance and size as measured by total assets in the population of banks not involved in mergers and acquisitions. Correlation is shown to be very low. For instance, when performance is measured by returns on equity, the Pearson correlation coefcient is equal to 2.4% whilst the Spearman correlation coecient is equal to 13.7%. In order to investigate the eect of size on the variation in performance determined by the merger or acquisition, we regress post-merger industryadjusted performance on pre-merger industry performance, the industry16

adjusted size of the acquirer, and the industry-adjusted size of the target (as outlined in equation 2). When performance is measured by returns on equity or cash-ow returns, the coecient is found to be negative and statistically signicant in most regressions [Table 12]. The empirical evidence hence suggests that performance attributable to a merger or acquisition is decreasing in the size of the acquirer. We now focus our attention on eciency measures. We rst compare the industry-adjusted performance of acquirers and targets [Table 13]. Both acquirers and targets are more ecient (both in prot and cost terms) than non-M&A banks9 , and the higher performance is statistically signicant (in line with the ndings on ROE and CFR). However, adjusted-values do not provide conrmation of the better performance of acquirers in comparison with targets when performance is measured by prot eciency. This result may be due to the higher standard deviation induced by the use of the control sample. The comparison of the eciency values of the combined bank emerging from the deal with the pre-values of the merging banks interestingly outlines improvements in cost eciency in the post-deal period, regardless of whether the base year prior to the deal refers to 3 or 6 years [Table 14 (Panel A)]. In each of the six years after the deal, cost eciency is higher than the cost eciency before the deal, and this happens in up to 80% of the cases (six years after the deal). Moreover, it emerges that improvements in cost eciency become more evident the longer the time after the deal, with a trend strictly monotonic (from +3.01% in year one after the deal to 5.10% in year six after the deal). By disentangling the sample into domestic and cross-border deals [Table 14 (Panel B and C)], the analysis suggests that the higher improvements in cost eciency are associated to domestic deals. The picture on the prot eciency side is however dierent. Prot efciency decreases in the post deal period in comparison with the pre-deal period [Table 14 (Panel A)]10 , and the decrease becomes more evident the
The industry-adjusted values show that banks involved in M&A operations are more (cost and prot) ecient than banks not involved in M&A (control sample) on average in any year (1991-2005) and in most of the countries under investigation. Further details can be found in a version of the paper available at http://www.fdic.gov/bank/analytical/cfr/Beccalli Frantz.pdf. 10 In order to make sure that our results are not driven by failure to control for size, we test the correlation between the (prot and cost) eciency and the (prot and cost) eciency measure normalised by equity. We nd large, positive, and signicant rank order correlation for each year before and after the deal. This correlation suggests that the eciency ordering generated by the unadjusted estimates is consistent with the eciency ordering generated by the normalised estimates. However the magnitude of the correlation coecient for prot eciency (about 0.75 signicant at 1% level) is lower than the
9

17

longer the number of years after the deal (as previously documented by the accounting protability measure). The average decrease in prot eciency varies between 1.17% (one year after the deal) and 5.33% (six years after the deal) when prot eciency post merger is compared with the weighted average of prot eciency for the target and acquirer in the six years prior to the deal. Interestingly, by distinguishing between domestic and cross-border operations [Table 14 (Panel B and C)], the decrease in prot eciency is particularly evident for cross-border operations; instead it does not emerge for domestic operations (as found for ROE). Note that with reference to cross-border M&As, the change in (cost and prot) eciency in the prevs. post-deal period is not aected by the location of the target either in a developed or in a developing country11 : the dierence in the change in performance for developed and developing countries is not statistically dierent from zero. Thus the location of the target in a developed vs. developing country does not have any impact on the change in performance. The previous ndings suggest that the impact of M&A operations on banks performance is negative on the prot eciency side and positive on the cost eciency side: M&A operations are associated with lower prot eciency and higher cost eciency. This could be due to the skimping on the resources devoted to underwriting and monitoring loans (Berger and DeYoung, 1997), which makes banks appear to be cost ecient in the short run because fewer operating expenses can support the same quality of loans12 . In addition, this seems to suggest that the improvements in cost eciency are transferred outside the bank, as bank revenues suer a decrease after the operation. It could be argued that cost benets are transferred to bank clients (and not to bank shareholders), especially in cross-border operations. The need to enter into new markets forces banks not to apply a price premium at least in the medium-term. To better investigate the above preliminary evidence, we disentangle the overall change in (cost and prot) eciency in order to isolate the variation specically determined by the M&A operation, by using the OLS regressions
magnitude of the correlation coecient for cost eciency (about 0.90 signicant at 1% level). Therefore we further investigate any scale bias eect on (prot and cost) eciency by testing the correlation between eciency and size (proxied by ln total assets) for the control sample. The magnitude of this correlation is very low (about 5% signicant at 1% level), which suggests that there is no failure to adjust for size the (prot and cost) eciency estimates. 11 As in much of the literature (Berger, 2007), the terms developed and developing countries are based on International Monetary Fund (IMF) classications: developed countries are high-income countries, developing countries are middle income or low income countries. 12 The authors would like to thank Allen Berger for suggesting such an interpretation.

18

previously outlined in equation (1). Several interesting results emerges for the overall sample of deals [Table 15 (Panel A)]. First, the explanatory power of the relationship is particularly high: by comparing the average of (both cost and prot) eciency in the 6 years after the deal with the average eciency in the 3 year before the deal, the R2 is much above 50%, a much higher value than the one traditionally found (e.g. as regard to cash ow return the R2 is 10% in Healy et al., 1992). Moreover, the decreasing trend over time in the values of the coecient suggests that there is a strong mean reversion trend in the industry-adjusted (cost and prot) eciency measures. This provides clear evidence of the highly competitive nature of the banking industry. Finally, the value of the intercept (a measure of the impact of the M&A operation itself) is positive and statistically signicant for cost eciency with respect to the overall sample both when the reference is to the 3 and 6 years prior to the deal. However, the value of the intercept in the regression involving prot eciency is not signicantly dierent from zero (Panel A). This would suggest that the M&A operation itself does have a positive impact on cost eciency, but does not have any (either positive or negative) impact on prot eciency. Furthermore, in order to investigate the size eect on the variation in performance determined by the M&A, we use the OLS regressions outlined in equation (2). The value of the intercept becomes positive and statistically signicant at 1% level, but the coecient is found to be negative and statistically signicant. This suggests that the larger the size of the acquirer, the lower the prot eciency associated with the M&A (Panel A). This surprising evidence leads us to further investigate the impact of the M&A operation itself by focusing on the level of geographical relatedness of the acquirer and target bank. To this aim, by distinguishing between domestic and cross-border operations, the analysis reveals that when the dependent variable is prot eciency, the value of the intercept is positive for domestic operations (Panel B) and negative for cross-border deals (Panel C). This implies that cross-border M&As have a negative impact on prot efciency, whereas domestic M&As have a positive impact on prot eciency. When the dependent variable is cost eciency, the value of the intercept is higher for domestic operations in comparison to cross-border operations. Overall, this suggests that for domestic deals the improvements in cost eciency and in prot eciency are due to the M&A operation itself, and not to the behaviour in X-eciency that would have occurred in absence of any M&A operation. For cross-border deals, instead, decreases in prot eciency occur because of the M&A operation itself, while the improvements in cost eciency are lower than those observed for domestic deals. Consequently, this evidence emphasizes the importance of geographical similarities in order 19

to achieve better post-M&A performance: geographical relatedness creates value. The potential determinants of the changes in cost and prot eciency due to M&A operations are proxied here by institutional/regulatory, bankspecic and deal-specic variables. Table 16 sets out their denitions and statistics. The rst category comprises freedom from government (an index measuring the incidence of all government expenditures and state-owned enterprises in the economy), regulatory quality (a measure of the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development) and banking concentration (total value of assets of the ve biggest banks). The second category includes the period in which the deal takes place (before and after year 2000), the method of payment used to regulate the operation (cash vs. equity), and the geographical nature of the operation (domestic vs. cross-border). The third category refers to the size of the banks involved in the operation (big, medium, and small measured on the basis of total assets), the focus of the banks involved in the so-called traditional banking (proxied by the amount of loans over total assets), and the degree of riskiness of the bank business (measured by the standard deviation of ROE). In order to identify the impact of these determinants on the changes in the eciency levels due to the M&A operation, we test equation (3) [Table 17]. With respect to the regulatory and institutional variables, the change in prot eciency (post vs. pre deal) is positively associated to the levels of freedom from government and regulatory quality characterising the home country of the target, whereas it is negatively associated to the same indexes qualifying the home country of the acquirer. Deals better able to create prot eciency are those in which acquiring banks direct their investments in countries with a better ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development and with lower government expenditures and state-owned enterprises. Instead, a higher concentration in the banking industry of the country of the acquirer has a negative impact on prot eciency. As regard to deal-specic conditions, cash payment has a negative impact on (prot and cost) eciency. The realisation of the M&A deal in the years 1994-1999 has a negative impact on (prot and cost) eciency, whereas when the deal occurs in the period 2000-2005 there is a negative eect on cost eciency only. With regard to structural bank-specic variables, a size qualied as medium (comprising all the banks in the 2nd tertile in terms of the natural logarithm of total assets) for the target in the pre-deal period results in a negative impact on prot eciency. Also a big size for the acquirer has a negative impact on prot eciency. A higher concentration of the acquirer 20

in the pre-deal period on traditional banking activities over the total bank activities (proxied by the proportion of loans over total assets) has a negative impact on both cost and prot eciency; whereas the impact is positive when the combined bank resulting from the operation has a higher concentration on traditional banking. Finally, the level of riskiness (proxied by the standard deviation of the ROE) of the activity of both the acquirer, the target, and the combined entity resulting form the M&A is always positively associated with the changes in prot and cost eciency. Finally, the evidence suggests that successful and unsuccessful deals dier with regard to several environmental and bank-characteristics [Table 18]13 . A higher likelihood of an unsuccessful M&A (measured in terms of prot eciency) is associated with a larger size of the acquirer, a larger equity of the acquirer and a higher risk (of both the acquirer and the target). Instead a larger diversication of the target is associated with a higher likelihood of an unsuccessful deal. Higher freedom from government and regulatory quality in the country of the acquirer are associated with a higher likelihood of an unsuccessful deal, whereas a higher regulatory quality in the country of the target is associated with a higher likelihood of a successful deal. A higher likelihood of an unsuccessful M&A (measured in terms of cost eciency) is associated with smaller size of the acquirer, a smaller capital of the acquirer, and a higher loan loss provision (of both the acquirer and the target).

Conclusions

This paper investigates whether M&A operations inuence the performance of banks. Using a sample of 714 deals involving EU acquirers and targets located throughout the world over the period 1991-2005, we investigate whether M&A operations are associated with improved performance (measured using both standard accounting ratios and cost and alternative prot X-eciency). Despite the extensive and ongoing consolidation process in the banking industry, we nd that M&A operations are associated with a slight deterioration in prot eciency, return on equity and cash ow returns, and a pronounced improvement in cost eciency in the 6 years after the deal (in comparison to the 3/6 years prior to the deal). Hence, the improvements in cost eciency appear to be transferred to bank clients rather than to bank shareholders.
Note that the dierence between successful and unsuccessful deals with regard to diversication and risk are not statistically dierent from zero for targets, whereas the quality of the loan portfolio is not statistically dierent from zero for acquirers. Similarly the method of payment, the cross-border nature of a deal, the location of the target in developing/developed country do not dier between successful and unsuccessful deals.
13

21

Interestingly, these changes in performance are directly attributable to the M&A operations and would not have occurred in their absence. Moreover, these changes exhibit a particularly negative trend for cross-border deals: in domestic deals, cost eciency improves more markedly than in cross-border deals, and prot eciency remains unchanged instead of diminishing. This highlights the importance of geographical relatedness in order to achieve better post-M&A performance. Finally, in the years before the M&A operation, target banks exhibit worse performance than acquiring banks in terms of prot eciency, protability accounting ratios, personnel expenses and operating costs. Besides, banks involved in M&A operations (both acquirers and targets) are more ecient and protable than their peers not involved in M&A operations. Furthermore, an important set of institutional, regulatory, bank-specic and deal-specic variables has a signicant inuence on the changes in cost and prot eciency. The management of acquiring banks should tend to direct investments to those countries that guarantee better regulatory quality together with higher freedom from government. Moreover, to achieve positive changes in eciency in the medium-term, transactions should be paid in equity (not in cash) and result in a combined bank with a higher focus on traditional banking activities. Finally, negative changes in prot eciency are associated with deals announced before year 2000, whereas more recent deals do not seem to determine any change in prot eciency. Finally, several environmental and bank-characteristics make a deal successful or unsuccessful. Successful M&As with respect to prot eciency are associated with a smaller size and equity of the acquirer, a lower risk (of both the acquirer and the target), a lower diversication of the target, a lower freedom from government and regulatory quality in the country of the acquirer, and a higher regulatory quality in the country of the target. Successful M&As with respect to cost eciency are associated with a larger size of the acquirer, a larger capital of the acquirer, and a lower loan loss provision (of both the acquirer and the target).

References
[1] Aigner, D.J., Lovell, C.A.K., Schmidt, P., 1977. Formulation and estimation of stochastic frontier production function models. Journal of Econometrics 6, 21-37. [2] Akhavein, J.D., Berger, A.N., D.B., Humphrey, D.B., 1997. The Eects

22

of Megamergers on Eciency and Prices: Evidence from a Bank Prot Function. Review of Industrial Organization 12, 95-139. [3] Altunbas, Y., Gardener, E.P.M., Molyneux, P., Moore, B., 2001. Eciency in European banking. European Economic Review 45, 1931-1955. [4] Altunbas, Y., Ibnez, D.M., 2004. Mergers and acquisitions and bank a performance in Europe. The role of strategic similarities. ECB Working paper n. 398, Frankfurt. [5] Altunbas, Y., Molyneux, P., Thornton, J., 1997. Big-Bank Mergers in Europe: An Analysis of the Cost Implications. Economica 64 (254), 317329. [6] Amel, D., Barnes, C., Panetta, F., Salleo, C., 2004. Consolidation and eciency in the nancial sector: A review of the international evidence. Journal of Banking and Finance 28(10), 2493-2519. [7] Battese, G.E., Coelli, T.J., 1995. A model for technical ineciency effects in a stochastic frontier production function for panel data. Empirical Economics 20, 325 -332. [8] Beccalli, E., 2004. Cross-Country Comparisons of Eciency: Evidence from the UK and Italian Investment Firms. Journal of Banking and Finance 28, 1363-1383. [9] Berger, A.N., 2007. International Comparisons of Banking Eciency. Financial Markets, Institutions & Instruments 16(3), 119-144. [10] Berger, A.N., 1998. The eciency eects of bank mergers and acquisitions: A preliminary look at the 1990s data, in Amihud, Y. and G. Miller (eds.), Bank mergers and acquisitions, Kluwer Academic Publishers: Boston, 79-111. [11] Berger, A.N., Demsetz, R.S., Strahan, P.E., 1999. The consolidation of the nancial services industry: Causes, consequences, and implications for the future. Journal of Banking and Finance 23, 135194. [12] Berger, A.N., DeYoung, R., 1997. Problem loans and cost eciency in commercial banks. Journal of Banking and Finance 21, 849-870. [13] Berger, A.N., DeYoung, R., Genay, H., Udell, G.F., 2000. The Globalization of Financial Institutions: Evidence from Cross-Border Banking 23

Performance. Brookings Wharton Papers on Financial Services, n. 3, 23 - 125. [14] Berger, A. N., Humphrey, D.B., 1997. Eciency of Financial Institutions: International Survey and Directions for Future Research. European Journal of Operational Research 98, 175-212. [15] Berger, A.N., Humphrey, D.B., 1992. Megamergers in Banking and the Use of Cost Eciency as an Antitrust Defense. Antitrust Bulletin 37, 541 600. [16] Berger, A.N., Leusner, J.H., Mingo J.J., 1997. The eciency of bank branches. Journal of Monetary Economics 40, 141-162. [17] Berger, A.N., Mester, L.J., 1997. Inside the black box: what explains dierences in the eciency of nancial institutions?. Journal of Banking and Finance 21, 895 - 947. [18] Coelli T.J., Prasada Rao D.S., Battese, G., 1998. An Introduction to Eciency and Productivity Analysis. Kluwer Academic Publishers, Norwell. [19] DeLong, G., DeYoung, R., 2007. Learning by Observing: Information Spillovers in the Execution and Valuation of Commercial Bank M&As. Journal of Finance 62, 181-216. [20] Demsetz, R.S., Strahan, P.E., 1997. Diversication, size, and risk at bank holding companies. Journal of Money, Credit, and Banking 29, 300 313. [21] DeYoung, R., 1997. Bank mergers, X-eciency, and the market for corporate control. Managerial Finance 23, 32-47. [22] Dietsch, M., Lozano-Vivas, A., 2000. How the environment determines banking eciency: A comparison between French and Spanish industries. Journal of Banking and Finance 24, 9851004. [23] Fama, E.F., French, K.R., 2000. Forecasting protability and earnings. Journal of Business 73, 161-175. [24] Focarelli D., Panetta F., 2003. Are mergers benecial to consumers? Evidence from the market for bank deposits. The American Economic Review 93(4), 1152-1172.

24

[25] Gallant, A.R., 1981. On the bias in exible functional forms and essentially unbiased form: The Fourier exible form. Journal of Econometrics 15, 211-245. [26] Gallant, A.R., 1982 Unbiased Determination of Production Technologies. Journal of Econometrics 20:285324. [27] Healy, P.M., Palepu, K.G., Ruback, R.S., 1992. Does corporate performance improve after mergers?, Journal of Financial Economics 31, 135-175. [28] Jagtiani, J. e Khanthavit, A., 1996. Scale and scope economies at large banks, including o balance sheet products and regulatory eects. Journal of Banking and Finance 20, 1271-1287. [29] Knapp, M., Gart, A., Chaudhry, M., 2006. The impact of mean reversion of bank protability on post-merger performance in the banking industry. Journal of Banking and Finance 30, 3503-3517. [30] Kwan, S.H., 2003. Operating performance of banks among Asian economies: An international and time series comparison. Journal of Banking and Finance 27, 471489. [31] Hughes J.P., Lang, W.W., Mester, L.J., Moon, C.-G., 1999. The dollars and sense of bank consolidation. Journal of Banking and Finance 23, 291324. [32] Huizinga, H.P., Nelissen, J.H.M., Vander Vennet, R., 2001. Eciency eects of bank mergers and acquisitions in Europe. Working Paper Ghent University n. 106. [33] Leibenstein, H., 1966. Allocative eciency vs. X-eciency. American Economic Review 56, 392-415. [34] McAllister, P., McManus, D., 1993. Resolving the scale eciency puzzle in banking. Journal of Banking and Finance 17, 389-405. [35] Mester, L.J., 1996. A study of bank eciency taking into account riskpreference. Journal of Banking and Finance 20, 1025-1045. [36] Mitchell, K., Onvural, N.M., 1996. Economies of scale and scope at large commercial banks: Evidence from the Fourier Flexible functional form. Journal of Money, Credit and Banking 28, 178-199.

25

[37] Peristiani, S., 1997. Do Mergers Improve the X-Eciency and Scale Eciency of U.S. Banks? Evidence from the 1980s. Journal of Money, Credit, and Banking 29(3), 326-337. [38] Pillo, S.J., 1996, Performance changes and shareholder wealth creation associated with mergers of publicly traded banking institutions. Journal of Money, Credit, and Banking 28, 294310. [39] Rhoades, S.A, 1998. The Eciency Eects of Bank Mergers: An Overview of Case Studies of Nine Mergers. Journal of Banking and Finance 22, 273-91. [40] Rhoades, S.A., 1993. Eciency eects of horizontal (in-market) bank mergers. Journal of Banking and Finance 17, 411422. [41] Rhoades, S.A., 1990. Billion Dollar Bank Acquisitions: a Note on the Performance Eects. Working Paper, Board of Governors of the Federal Reserve System, Washington, DC. [42] Sealey C., Lindley J.T. Inputs, outputs and a theory of production and cost at depository nancial institutions. Journal of Finance 32(1977), 1251-66. [43] Trichet, J.C. Towards the review of the Lamfalussy approach market developments, supervisory challenges and institutional arrangements. First CEBS Conference, London, 2007. [44] Vander Vennet, R., 1996. The eect of mergers and acquisitions on the eciency and protability of EC credit institutions. Journal of Banking and Finance 20 (1996) 1531-1558. [45] Vander Vennet, R., 2002. Cross-border mergers in European banking and bank eciency. Working Paper Ghent University n. 152.

26

Table 1: Number of M&A deals (by country and by year); 1991 -2005
1991 Austria Belgium Denmark Finland France Germany Greece Hungary Iceland Ireland Italy Luxembourg Netherlands Norway Poland Portugal Spain Sweden Switzerland Turkey UK Total 1 22 3 27 1 29 3 63 3 54 5 48 5 71 2 60 4 87 2 85 1 52 1 37 1 35 2 42 2 4 4 6 1 2 5 6 2 6 2 3 1 3 2 4 2 9 2 1 2 2 1 3 3 1 3 16 2 22 8 14 1 1 2 2 1 8 1 2 11 2 1 6 3 11 8 1 2 11 12 2 1 6 3 6 1 1 4 12 2 3 3 22 28 1 3 20 1 2 16 11 1 1 1 1 1 5 1 2 1 3 2 3 1 5 2 1 11 2 7 5 1 1 1 9 3 2 1 17 10 8 7 9 5 7 15 6 8 2 5 9 6 4 11 3 9 2 1 1 3 1 2 2 1 2 3 4 4 10 2 1 1992 1993 1994 1 1995 2 1996 1 1997 4 4 2 1998 2 4 2 1999 5 2 2 2000 1 1 3 4 2 2 1 2001 2002 2003 2004 1 2 2 2005 Total 17 20 18 2 109 80 16 5 6 2 178 8 22 12 17 27 93 19 28 1 34 714

27

Table 2: Number of cross-border M&A deals (by country); 1991-2005

Home country acquirer Home country target AU Argentina Australia Austria Belgium Brazil Bulgaria Canada Chile Colombia Croatia Czech Republic Denmark Estonia Finland France Germany Greece Hungary India Ireland Italy Lebanon Luxembourg Mexico Morocco Netherlands Norway Poland Portugal Romania Slovak Rep Slovenia South Africa South Korea Spain Sweden Switzerland Thailand Turkey United Kingdom United States Venezuela 1 1 1 1 2 2 1 4 1 3 4 3 1 1 2 1 1 2 1 1 1 3 1 1 1 1 1 3 1 3 3 1 2 4 3 4 1 1 1 5 1 2 1 1 3 3 1 1 1 4 13 2 1 1 3 1 1 4 3 7 1 2 3 2 1 1 3 1 1 5 1 2 13 5 4 4 2 1 2 4 5 3 2 4 1 2 1 1 1 4 2 3 2 2 2 1 1 2 1 2 1 1 3 1 1 1 1 2 6 3 1 4 3 1 1 1 1 2 2 1 1 1 1 1 5 6 1 4 1 1 1 4 1 5 1 1 BE DE FR 1 GE GR HU IS IR IT 1 LU NE 1 PL PO SP 12 1 1 2 SE CH TR UK Total 15 3 9 4 13 4 2 6 7 3 8 8 3 1 25 13 5 13 1 3 20 1 5 13 3 2 6 33 9 8 6 1 4 4 20 2 5 1 2 11 15 3

Total 7 14 8 47 51 5 5 4 1 22 8 21 6 10 66 16 8 1 20 320 AU: Austria; BE : Belgium; DE: Denmark; FR: France; GE: Germany; GR: Greece; HU: Hungary; IS: Iceland; IR: Ireland; IT: Italy; LU: Luxembourg; NE: Netherlands; PL: Poland; PO: Portugal; SP: Spain; SE: Sweden; CH: Switzerland; TR: Turkey; UK: United Kingdom

28

Table 3: Number of banks in the control sample (by country and by year)
Year Country Argentina Australia Austria Belgium Brazil Canada Chile Colombia Denmark Finland France Germany Greece Hungary India Ireland Italy Lebanon Luxembourg Mexico Morocco Netherlands Norway Poland Portugal Romania Slovenia South Africa South Korea Spain Sweden Switzerland Thailand Turkey United Kingdom United States Venezuela Total 1 9 21 1 103 1 9 23 1 119 8 3 4 5 5 1 8 3 5 6 3 7 3 8 1 1 9 214 1 336 6 4 8 4 10 1 1 10 223 1 371 1 7 4 8 4 11 2 1 13 237 1 414 1 7 4 8 4 13 2 1 16 248 2 443 1 10 5 9 4 13 4 1 18 262 3 483 1 12 5 9 4 13 7 1 20 303 3 539 1 13 4 9 5 14 10 1 23 344 3 599 2 4 4 4 4 1 1 3 2 1 1 1 3 3 2 3 3 3 3 10 1 1 1 1 3 4 3 11 1 1 1 1 3 5 3 15 2 1 1 1 3 6 3 16 2 2 1 1 5 7 2 4 3 18 2 3 2 1 5 7 5 5 4 20 2 3 2 1 5 8 5 5 4 21 2 3 3 1 5 8 6 5 4 25 2 3 3 3 5 10 7 5 1 1 13 5 9 6 18 13 1 25 359 5 649 1 6 2 9 4 7 2 12 4 1 1 8 3 15 5 2 1 11 3 16 6 5 2 14 3 17 6 5 2 14 3 17 6 6 2 14 3 17 6 8 1 4 3 5 3 6 4 6 10 1 8 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 1 8 1 9 1 2 10 1 3 10 6 1 8 10 4 3 10 7 1 10 11 4 5 10 8 5 9 11 5 5 10 9 5 10 12 7 1 14 3 17 4 8 1 5 10 8 4 9 12 6 1 14 3 25 5 10 2 5 11 7 5 10 12 7 2 14 4 29 7 10 3 5 12 6 5 14 14 7 2 14 4 31 8 10 3 13 4 27 2 2 5 3 5 11 8 5 1 1 13 7 9 7 16 20 2 27 391 7 721 5 14 6 5 16 16 7 3 14 4 31 8 11 3 19 5 28 2 2 5 3 5 12 8 5 1 1 13 8 9 9 16 20 6 29 468 7 824 1 12 9 9 9 17 20 9 31 488 7 854 1 12 9 9 9 14 22 9 31 498 7 860 11 18 7 10 484 7 648 4 8 1 6 14 6 5 16 15 7 4 15 4 31 8 11 3 16 5 29 2 2 6 3 6 14 9 5 6 14 6 5 16 16 7 4 15 4 31 6 11 3 17 4 29 1 2 5 3 6 14 9 5 5 3 4 1 1 1 18 1 1 1 15 15 7 3 3 6 12 1 Total 58 162 72 41 154 164 69 20 177 45 302 84 99 29 83 51 256 22 26 42 27 62 111 59 62 3 10 138 76 120 74 183 140 43 280 4563* 56 7963

*When estimating efficiency, in order to avoid the bias induced by a disproportionate presence of US banks in the frontier of the control sample, we have randomly selected and included only a sub-sample of 45 US banks.

29

Table 4: Comparison of unadjusted values of efficiency and return on equity for acquirer and target prior to the M&A deal
Panel A. Acquirers. Unadjusted values for cost efficiency, profit efficiency and returns on equity
N B1 B2 B3 B4 B5 B6 721 674 612 529 464 410 N B1 B2 B3 B4 B5 B6 222 212 187 155 137 120 N B1 B2 B3 B4 B5 B6 209 194 168 130 106 95 Cost efficiency Mean Std. Dev 0.8042 0.1019 0.8038 0.0935 0.7953 0.1002 0.7940 0.1012 0.7879 0.1021 0.7809 0.1039 Cost efficiency Mean Std. Dev 0.7928 0.1177 0.7910 0.1165 0.7896 0.1198 0.7849 0.1076 0.7796 0.1127 0.7734 0.1214 Cost efficiency Mean Std. Dev 0.0091 0.1376 0.0129 0.1390 0.0094 0.1345 0.0084 0.1369 0.0056 0.1328 -0.0021 0.1480 Profit efficiency Mean Std. Dev 0.7963 0.0780 0.7997 0.0764 0.7999 0.0758 0.8018 0.0688 0.8016 0.0665 0.8106 0.0614 Profit efficiency Mean Std. Dev 0.7705 0.1318 0.7700 0.1281 0.7821 0.1267 0.7835 0.1161 0.7765 0.1287 0.7867 0.1212 Profit efficiency Mean Std. Dev 0.0132* 0.1392 0.0244** 0.1325 0.0190* 0.1395 0.0161 0.1242 0.0312** 0.1322 0.0322** 0.1284 ROE Mean Std. Dev 0.1278 0.0772 0.1209 0.0644 0.1204 0.0631 0.1138 0.0656 0.1164 0.0760 NTB Mean Std. Dev 0.6404 3.0359 0.5741 1.0761 0.7479 1.4417 0.8394 1.2418 1.1760 1.8698 1.1755 1.4886 NTB Mean Std. Dev 0.3494 0.9433 0.3474 0.4488 0.4241 0.5923 0.3936 0.5513 0.4810 0.7428 0.4291 0.3298 NTB Mean Std. Dev 0.1032** 0.5620 0.1030* 0.5616 0.0888 0.9359 0.3355 1.1665 0.5244 2.0823 0.0498 0.1307 NM Mean 0.0733 0.0660 0.0589 0.0556 0.0555 0.0585 NM Mean 0.0631 0.0599 0.0648 0.0560 0.0404 0.0315 Std. Dev 0.1260 0.0746 0.0677 0.0781 0.0752 0.1075 Std. Dev 0.0640 0.0485 0.0565 0.0637 0.0549 0.0501 TOER Mean Std. Dev 0.0228 0.0092 0.0225 0.0089 0.0221 0.0086 0.0223 0.0086 0.0233 0.0088 0.0225 0.0081 TOER Mean Std. Dev 0.0268 0.0190 0.0264 0.0151 0.0257 0.0137 0.0264 0.0136 0.0278 0.0153 0.0275 0.0138 TOER Mean Std. Dev -0.0065*** 0.0207 -0.0050*** 0.0147 -0.0047*** 0.0137 -0.0056*** 0.0131 -0.0060*** 0.0165 -0.0056*** 0.0152 Mean 0.0125 0.0124 0.0123 0.0123 0.0127 0.0123 PER Std. Dev 0.0050 0.0051 0.0049 0.0045 0.0048 0.0047 PER Std. Dev 0.0115 0.0105 0.0095 0.0101 0.0093 0.0110

Panel B. Targets. Unadjusted values for cost efficiency, profit efficiency and determinants
ROE Mean Std. Dev 0.1039 0.1396 0.1109 0.1101 0.1281 0.1136 0.1197 0.1415 0.1120 0.1455 Mean 0.0158 0.0155 0.0148 0.0148 0.0150 0.0159

Panel C. Acquirers versus targets. Unadjusted values for cost efficiency, profit efficiency and determinants
ROE Mean 0.0505*** 0.0236*** 0.0118 0.0132 0.0238** 0.0661***
Std. Dev

0.2217 0.0987 0.1101 0.1383 0.1228 0.2381

Mean 0.0226* 0.0114 -0.0048 0.0033 0.0123 0.0209*

NM Std. Dev 0.13126 0.0827 0.0622 0.0776 0.0700 0.0711

PER Mean Std. Dev -0.0046*** 0.0123 -0.0044*** 0.0109 -0.0037*** 0.0097 -0.0040*** 0.0112 -0.0040*** 0.0105 -0.0047*** 0.0128

Return on Equity (ROE) = Net income/Total Equity (end of the year); Non Traditional Banking (NTB) = OBS/Total assets; Net margin (NM) = Net income/Revenues (= Interest income + Commission income + Trading income); Total Operating Expense Ratio (TOER) = Total non-interest operating expense/Total assets; Personnel expense ratio (PER) = Personnel costs/Total assets. ***, **, * T-test respectively statistically significant at 1%, 5% and 10%. N: number of observations.

30

Table 5: Comparison of return on equity before and after the deal

Panel A. Acquirer. Industry-adjusted values for ROE(BY) before the deal (adjustment: by year, country and type of activity) Year Acquirer Acquirer Industry-Adjusted Median ROE (%) N Median ROE (%) N % + Z Score B1 12.2 691 2.07 618 63.1 8.82*** B2 11.6 641 2.25 566 69.3 10.21*** B3 11.9 553 2.49 501 72.9 11.04*** B4 11.1 485 2.28 442 74.4 10.60*** B5 10.6 426 2.57 396 73.5 10.60*** Panel B. Target. Industry-adjusted values for ROE(BY) before the deal (adjustment: by year, country and type of activity) Year Target Target Industry-Adjusted Median ROE (%) N Median ROE (%) N % + Wilcoxon Test Z Score B1 11.0 272 1.56 224 60.3 2.21** B2 10.1 254 1.22 206 57.8 1.96** B3 10.0 215 0.54 169 55.0 1.99* B4 9.7 177 1.96 135 60.7 1.87* B5 9.1 151 1.11 116 57.8 1.75* Panel C. Acquirer Target. Industry-adjusted values for ROE(BY) before the deal (adjustment: by year, country and type of activity) Year Industry-Adjusted ROE (%) Wilcoxon Test Acquirer Target Difference N % + Z Score B1 2.07 1.56 0.02 169 50.3 1.10 B2 2.25 1.22 0.00 157 48.4 1.08 B3 2.49 0.54 0.57 119 52.9 1.73* B4 2.28 1.96 1.14 91 53.8 2.12** B5 2.57 1.11 2.41 78 59.0 2.96*** Panel D. Post values for the combined bank. Adjusted values for ROE(BY) (adjustment: by year, country and type of activity) Year Combined bank Combined bank Industry-Adjusted Median ROE (%) N Median ROE (%) N %+ Z Score A1 12.4 675 2.09 111 65.8 3.03*** A2 11.6 663 0.80 94 58.5 1.60 A3 11.5 629 0.83 83 59.0 1.27 A4 11.6 565 -0.34 68 48.5 0.99 A5 12.4 493 0.21 51 51.0 0.35 Panel E. Post values for the combined bank vs. Pre values of the merging banks. Adjusted values for ROE(BY) (adjustment: by year, country and type of activity) Pre-Merger Post-Merger Post versus Pre Wilcoxon Test Years Median ROE(%) N Years Median ROE(%) N Change %+ Z Score (N) B2B1 2.09 186 A1A2 1.80 117 0.00 45.1 0.02 (102) B3B1 2.00 189 A1A3 1.33 118 -1.00 39.4 2.00** (104) B5B1 1.94 191 A1A5 0.99 119 -2.50 31.7 3.26*** (104)

ROE = Net income / Total Equity (beginning of the year). N: number of observations. % of positive cases under the Wilcoxon test. ***, **, * statistical significance at 1%, 5% and 10%.

Table 6: Regression of post-merger on pre-merger industry adjusted mean ROE

Panel A. Regression of Annual Post-Merger Industry-Adjusted ROE on Average Pre-Merger Industry-Adjusted ROE R R2 Adj.R2 N A1, B3B1 -0.014 0.614*** 0.331 0.109 0.095 62 (0.010) (0.226) A2, B3B1 -0.034** 1.108*** 0.477 0.227 0.212 51 (0.013) (0.292) A3, B3B1 -0.032*** 0.214 0.118 0.014 -0.012 40 (0.012) (0.292) A4, B3B1 -0.025** 0.354 0.216 0.047 0.013 30 (0.010) (0.302) A5, B3B1 -0.035*** -1.242*** 0.676 0.456 0.428 21 (0.010) (0.311) Panel B: Regression of Average Post-Merger Industry-Adjusted ROE on Average Pre-Merger Industry-Adjusted ROE A1A2, B3B1 -0.024** 1.055*** 0.529 0.280 0.265 51 (0.011) (0.242) A1A3, B3B1 -0.026** 0.508 0.320 0.103 0.079 40 (0.010) (0.244)** AdjROEAi denotes the industry-adjusted ROE of the combined firm in the ith year following the acquisition.AdjROEA1Ai denotes the average industryadjusted ROE of the combined firm from the first to the ith year following the acquisition. AdjROEB1B3 denotes the average industry-adjusted ROE of the combined firm from the third to the first year prior to the acquisition.

31

Table 7: Domestic versus cross-border M&A deals

Panel A. Domestic vs. cross-border deals. Acquirer Target. Industry-adjusted values for ROE(BY) before the deal Domestic M&A Deals Cross-Border M&A Deals Mean ROE Wilcoxon Test Mean ROE Wilcoxon Test N. of years Acquirer Target Acquirer vs Target Acquirer Target Acquirer vs Target before the deal % % N %+ Z % % N %+ Z B1 2.63 -0.66 84 54.8 2.71*** -0.01 1.14 85 44.7 1.26 B2 1.51 -0.48 71 52.6 2.11** 0.85 2.62 79 37.2 1.71* B3 1.32 0.18 56 60.7 2.59*** 2.33 4.03 63 47.6 0.49 B4 1.57 -1.95 41 56.1 3.55*** 1.83 4.49 50 56.0 0.44 B5 2.50 1.69 37 70.3 3.55*** 2.45 0.76 41 48.8 0.32 Panel B. Domestic vs. cross-border deals. Post values for the combined bank vs. Pre values of the merging banks. Adjusted values for ROE(BY) Domestic M&A Deals Cross-Border M&A Deals N. of years Mean ROE Wilcoxon Test Mean ROE Wilcoxon Test before the deal N ROE (%) N Change in %+ N % N Change in %+ ROE (%) (Z) ROE (%) (Z) B3B1 53 0.31 99 0.52 A1 50 -1.61 27 -2.19 44.4 273 -0.77 69 0.06 0.79 (1.18) A2 38 -1.04 22 -2.21 54.5. 281 -1.84 64 -2.27 26.6 (0.47) 2.88*** A3 36 -0.27 19 -1.42 52.6 265 -1.28 56 -1.82 32.1 (0.44) 2.87*** A4 31 -1.37 14 -2.29 42.9 231 -1.06 45 -1.43 26.7 (0.22) 2.53** A5 25 5.36 11 1.29 54.5 206 -2.10 34 -3.83 20.6 (0.62) 3.22*** A1A2 88 -0.60 22 -2.47 40.9 302 -1.05 72 -0.98 37.5 (1.45) 1.90* A1A3 74 -0.59 19 -1.19 57.9 314 -1.07 73 -1.14 38.4 (0.73) 1.90* A1A5 45 +0.05 21 1.15 72.7 336 -1.35 73 -1.86 28.8 (1.16) 3.00*** In this table, in the interest of concision, ROE refers to mean industry-adjusted Return on Equity, where equity is measured at the beginning of the year. In the same spirit, Change in ROE refers to the difference between the mean industry-adjusted Return on Equity in some year following the acquisition and the mean industry-adjusted Return on Equity in the 3 years before the acquisition.

Table 8: Comparison of net margin and asset turnover before and after the deal
Panel A. Acquirer. Industry-adjusted values for NBM and BAT before the deal (adjustment: by year, country and type of activity) NBM BAT N. of years NBM Adjusted NBM Wilcoxon Test BAT Adjusted BAT Wilcoxon Test before the Median N Median N %+ %Z Median N Median N %+ %Z deal (%) (%) (%) (%) B1 18.0 746 0.00 667 46.3 41.4 3.15*** 2.27 642 0.15 533 55.5 44.5 0.02 B2 17.2 720 0.00 645 48.7 41.1 3.28*** 2.20 592 0.12 505 56.4 43.6 0.41 B3 17.0 649 0.00 578 45.8 42.4 2.40** 2.17 521 0.04 461 53.4 46.6 1.43 B4 15.9 557 0.00 504 49.0 38.3 3.15*** 2.17 452 0.02 407 50.4 49.6 1.00 B5 15.1 493 0.00 453 47.2 40.4 2.01** 2.18 404 0.04 375 51.7 48.3 0.18 Panel B. Target. Industry-adjusted values for NBM and BAT before the deal (adjustment: by year, country and type of activity) NBM BAT N. of years NBM Adjusted NBM Wilcoxon Test BAT Adjusted BAT Wilcoxon Test before the Median N Median N %+ %Z Median N Median N %+ %Z deal (%) (%) (%) (%) B1 18.5 300 0.77 251 40.2 54.2 2.40** 2.07 235 -0.12 180 47.2 52.8 2.65*** B2 17.2 286 0.00 236 41.1 51.3 1.75* 2.06 217 -0.19 168 45.2 54.8 3.22*** B3 16.9 256 -0.57 206 39.3 53.9 1.35 2.13 183 -0.15 140 43.6 56.4 2.75*** B4 16.0 212 -0.30 164 42.7 50.0 1.58 1.97 160 -0.05 119 45.4 54.6 2.09** B5 14.9 177 -1.96 134 39.6 56.7 2.50** 2.02 134 -0.11 97 47.4 52.6 1.18 Panel C. Post values for the combined bank vs. Pre values of the merging banks. Adjusted values for NBM and BAT (adjustment: by year, country and type of activity) Pre-Merger Post-Merger Post versus Pre Wilcoxon Tesrs Year Year N NBM NBM N BAT N NBM N BAT N NBM NBM BAT BAT N BAT Median Median Median Median %+ Z %+ Z (%) (%) (%) (%) (%-) (%-) B2B1 -0.05 189 0.01 149 A1A2 -0.11 119 -0.04 97 53.3 0.69 34.1 0.94 105 (40.0) (39.0) 82 B3B1 -0.36 194 0.01 147 A1A3 0.00 120 -0.02 97 50.9 0.07 31.3 1.48 106 (39.6) (46.3) 80 B5B1 -0.36 194 0.04 141 A1A5 -0.08 121 -0.09 89 50.0 0.03 21.9 1.83* 106 (43.4) (38.4) 73 NBM = Net Income / Net Revenue. BAT = Net Revenue / [Total Assets + Off Balance Sheet Assets (beginning of the year)]. % +: % of positive cases under the Wilcoxon test. % -: % of negative cases under the Wilcoxon test. ***, **, * statistical significance at 1%, 5% and 10%.

32

Table 9: Comparison of ratios of personnel and other administrative expenses over revenue before and after the deal
Panel A. Acquirer. Industry-adjusted values for PEX and AOE before the deal (adjustment: by year, country and type of activity) N. of years PEX OAE before the deal PEX Adjusted PEX Wilcoxon Test OAE Adjusted OAE Wilcoxon Test Median N Median N %+ %Z Median N Median N %+ %Z (%) (%) (%) (%) B1 15.9 730 -1.18 651 24.1 60.8 10.2*** 9.63 694 -0.20 620 30.0 54.8 6.47*** B2 15.6 704 -1.26 630 25.9 61.9 10.2*** 9.32 674 -0.21 603 31.5 52.9 6.30*** B3 14.8 638 -1.46 567 23.3 63.3 10.1*** 8.60 610 -0.31 542 27.7 55.4 6.69*** B4 14.6 549 -1.31 497 23.3 63.0 9.36*** 8.48 526 -0.36 476 28.6 56.3 6.26*** B5 14.1 485 -1.47 444 25.9 63.3 8.97*** 8.36 465 -0.25 426 30.0 56.1 6.37*** Panel B. Target. Industry-adjusted values for PEX and OAE before the deal (adjustment: by year, country and type of activity) N. of years PEX OAE before the deal PEX Adjusted PEX Wilcoxon Test OAE Adjusted OAE Wilcoxon Test Median N Median N %+ %Z Median N Median N %+ %Z (%) (%) (%) (%) B1 16.4 292 -0.34 243 36.6 52.3 0.93 10.50 217 0.00 185 45.9 49.2 0.46 B2 15.9 275 -0.47 225 37.8 53.8 1.54 9.90 215 0.00 181 43.1 50.3 0.72 B3 15.4 241 -0.49 193 34.7 54.9 2.28** 9.60 195 -0.07 162 42.0 50.6 1.30 B4 15.6 202 -0.49 155 34.8 56.1 1.81* 9.85 167 0.00 133 47.4 46.6 0.61 B5 14.7 173 -0.90 130 36.2 57.7 2.74*** 8.75 147 -0.05 114 40.4 51.8 1.86* Panel C. Post values for the combined bank vs. Pre values of the merging banks. Adjusted values for PEX and OAE (adjustment: by year, country and type of activity) Pre-Merger Post-Merger Post versus Pre Wilcoxon Tesrs N. of years PEX N OAE N Years PEX N OAE N PEX PEX OAE OAE N PEX before the deal Median Median Median Median %+ Z %+ Z N OAE (%) (%) (%) (%) (%-) (%-) B2B1 -1.68 176 -0.39 145 A1A2 -1.15 110 -0.07 89 54.7 1.86* 58.2 2.00** 95 (45.3) (41.8) 79 B3B1 -1.69 180 -0.51 148 A1A3 -1.07 111 -0.01 90 57.3 2.40** 61.3 2.18** 96 (40.6) (37.5) 80 B5B1 -1.92 183 -0.70 150 A1A5 -1.28 112 0.00 91 62.5. 2.96*** 62.5. 2.70*** 96 (36.5) (36.3) 80 PEX = Personnel Expense / Revenue. OAE = Other Administrative Expense / Revenue. % +: % of positive cases under the Wilcoxon test. % -: % of negative cases under the Wilcoxon test. ***, **, * statistical significance at 1%, 5% and 10%.

Table 10: Regressions of industry-adjusted post-merger accounting performance on industryadjusted pre-merger accounting performance
AdjPerM & A , post = + AdjPerM & A , pre +

Post, Pre

R2

AdjR2

N 103 104 104 104 93 92 94 94

Post, Pre

R2

AdjR2

Panel A: Net Margins (NBM) A1A2, B3B1 -0.070*** -1.499*** 0.389 0.383 (0.023) (0.187) A1A3, B3B1 -0.061*** -1.444*** 0.415 0.410 (0.020) (0.170) A1A4, B3B1 -0.080*** -1.733*** 0.566 0.561 (0.018) (0.150) A1A5, B3B1 -0.084*** -1.719*** 0.562 0.557 (0.018) (0.150) Panel C: Personnel Expense over Revenue (PEX) A1A2, B3B1 0.008 1.394*** 0.648 0.644 (0.011) (0.108) A1A3, B3B1 -0.019** 0.394*** 0.095 0.085 (0.009) (0.128) A1A4, B3B1 -0.007 1.316*** 0.544 0.539 (0.012) (0.126) A1A5, B3B1 -0.018 1.243*** 0.439 0.433 (0.014) (0.146) Panel E: Cash Flow Returns CFR (BY) A1A2, B3B1

Panel B: Asset Turnover (BAT) A1A2, B3B1 -0.008*** 0.264*** 0.314 0.305 78 (0.002) (0.045) A1A3, B3B1 -0.009*** 0.233*** 0.259 0.249 79 (0.002) (0.045) A1A4, B3B1 -0.009*** 0.215*** 0.229 0.219 79 (0.002) (0.045) A1A5, B3B1 -0.009*** 0.203*** 0.215 0.205 79 (0.002) (0.044) Panel D: Other Administrative Expenses over Revenue (OAE) A1A2, B3B1 0.019** 1.574*** 0.801 0.799 77 (0.010) (0.090) A1A3, B3B1 0.004 1.138*** 0.735 0.731 78 (0.008) (0.078) A1A4, B3B1 -0.008 0.919*** 0.627 0.622 78 (0.009) (0.081) A1A5, B3B1 -0.018* 0.787*** 0.496 0.490 78 (0.010) (0.091) Panel F: Cash-Flow Returns CFR(AVG)

-0.113 0.507* 0.062 0.040 45 A1A2, B3B1 -0.401*** 0.848** 0.087 0.066 45 (0.070) (0.300) (0.046) (0.420) A1A3, B3B1 -0.116* 0.514* 0.078 0.057 45 A1A3, B3B1 -0.378*** 0.985*** 0.141 0.125 45 (0.063) (0.269) (0.040) (0.364) A1A4, B3B1 -0.132** 0.493* 0.073 0.052 45 A1A4, B3B1 -0.372*** 1.008*** 0.155 0.135 45 (0.062) (0.268) (0.039) (0.359) A1A5, B3B1 -0.136** 0.478* 0.073 0.052 45 A1A5, B3B1 -0.372*** 0.992*** 0.149 0.129 45 (0.060) (0.259) (0.039) (0.361) Adj PerA1Ai denotes the average industry-adjusted performance of the combined bank from the first to the ith year following the acquisition. Adj PERB1B3 denotes the average industry-adjusted performance of the combined bank from the third to the first year prior to the acquisition. CFR (BY) = (Net Revenue Personnel Expense Other Administrative Expense)/Market Value (Beginning of the year). CFR (AVG) = (Net Revenue Personnel Expense Other Administrative Expense)/Market Value (Average over the year).

33

Table 11: Comparison of cash-flow returns before and after the deal
Panel A. Industry-adjusted values for CFR before the deal (adjustment: by year, country and type of activity) N. of years Acquirer Acquirer Industry-Adjusted before the Mean CFR N Mean N % + % - Z Score deal (%) CFR (%) B1 21.2 513 1.46 467 53.3 45.4 2.04** B2 23.1 484 4.27 433 60.7 37.4 4.89*** B3 23.5 418 4.84 382 63.6 35.9 5.62*** B4 24.7 360 6.13 332 67.8 30.4 6.90*** B5 27.1 316 4.83 292 69.2 29.5 7.28*** B2B1 22.7 540 3.29 494 55.8 43.8 3.85*** B3B1 24.2 551 4.98 507 62.8 37.2 4.73*** B5B1 26.8 561 6.86 517 64.3 35.7 6.70*** Panel C. Acquirer Target. Industry-adjusted values for CFR before the deal (adjustment: by year, country and type of activity) N. of years Mean Industry Acquirer versus Target (Wilcoxon before the Adjusted CFR (%) Test)
deal

Acquirer B1 B2 B3 B4 B5 B2B1 B3B1 B5B1 1.46 4.27 4.84 6.13 4.83 3.29 4.98 6.86

Target -1.43 -1.10 0.41 -2.99 -2.56 -0.58 0.30 0.36

N 72 73 49 38 30 85 85 85

%+ 48.6 47.9 42.9 44.7 30.0 50.0 45.9 43.5

%29.2 34.2 34.7 28.9 46.7 30.5 36.5 38.8

Z Score 2.37** 2.53** 1.05 1.64 0.09 2.35** 1.55 1.61

Panel B. Target. Industry-adjusted values for CFR before the deal (adjustment: by year, country and type of activity) N. of years Target Target Industry-Adjusted before the Mean CFR N Mean N % + % Z deal (%) CFR (%) Score B1 18.6 113 -1.43 100 53.0 47.0 0.32 B2 20.1 112 -1.10 98 45.9 53.1 0.20 B3 24.5 92 0.41 77 50.6 45.5 0.48 B4 25.8 79 -2.99 65 63.1 36.9 1.08 B5 27.3 63 -2.56 52 55.8 44.2 0.62 B2B1 18.2 101 -3.47 86 46.5 53.5 1.32 B3B1 18.6 84 -3.36 68 47.1 52.9 1.23 B5B1 21.6 57 -5.14 45 53.3 46.7 0.80 Panel D. Post values for the combined bank vs. Pre values of the merging banks N. of years Mean Industry Change in MeanWilcoxon Test after the Adjusted CFR Industry deal Adjusted CFR N (%) N (%) %+ %- Z Score

B3B1 82 2.7 A1 82 -16.5 56 -15.1 28.6 71.4 3.30*** A2 68 -6.1 44 -5.9 27.3 72.7 2.42** A3 58 -4.4 38 -10.5 23.7 76.3 3.75*** A4 50 -13.3 34 -20.9 20.6 79.4 4.35*** A5 39 -10.2 24 -20.1 8.3 91.7 3.91*** A1A2 85 -14.1 57 -14.8 24.6 75.4 3.77*** A1A3 85 -13.1 57 -14.2 22.8 77.2 4.43*** A1A5 85 -13.2 57 -15.7 19.3 80.7 4.58*** % +: % of positive cases under the Wilcoxon test. % -: % of negative cases under the Wilcoxon test. ***, **, * statistical significance at 1%, 5% and 10%.

Table 12: Regression of industry-adjusted post-merger accounting performance on industryadjusted pre-merger accounting performance and industry adjusted acquirer and target size
AdjPerM & A, post = + AdjPerM & A, pre + SizeA, pre + SizeT , pre +
Post, Pre A1,B1 A1A2, B3B1 A1A3, B3B1 A1A4, B3B1 A1A5, B3B1 A1,B1 A1A2, B3B1 A1A3, B3B1 A1A4, B3B1 A1A5, B3B1 Panel A: Return on Equity ROE (BY) 1 2 0.004 0.275*** -0.013** 0.001 (0.010) (0.091) (0.006) (0.004) 0.004 0.204*** -0.011** 0.005 (0.007) (0.068) (0.004) (0.003) 0.000 0.208*** -0.010*** 0.003 (0.007) (0.061) (0.004) (0.003) -0.003 0.209*** -0.010** 0.002 (0.006) (0.060) (0.004) (0.003) -0.008 0.201*** -0.005 0.003 (0.006) (0.057) (0.004) (0.003) Panel C: Cash Flow Returns CFR (BY) -0.167 1.096*** -0.033 0.030 (0.144) (0.327) (0.067) (0.055) -0.100 0.859*** -0.059 0.028 (0.065) (0.166) (0.039) (0.033) -0.091* 0.878*** -0.063** 0.028 (0.052) (0.131) (0.031) (0.026) -0.104** 0.883** -0.060* 0.024 (0.052) (0.132) (0.031) (0.026) -0.109** 0.879*** -0.059* 0.021 (0.050) (0.126) (0.030) (0.025) AdJ.R2 0.064 0.074 0.086 0.085 0.079 0.144 0.333 0.466 0.460 0.480 N 123 132 133 133 133 54 56 56 56 56 Panel B: Return on Equity ROE (AVG) 1 2 0.008* 0.165*** -0.007** -0.002 (0.005) (0.044) (0.003) (0.002) 0.009** 0.266** -0.006** 0.001 (0.004) (0.058) (0.003) (0.002) 0.009** 0.260*** -0.005** 0.001 (0.004) (0.053) (0.003) (0.002) 0.008** 0.237*** -0.004* 0.000 (0.003) (0.050) (0.002) (0.002) 0.005 0.196*** -0.001 0.001 (0.003) (0.050) (0.002) (0.002) Panel D: Cash Flow Returns CFR (AVG) -0.436*** 1.520*** -0.073 0.028 (0.094) (0.264) (0.055) (0.045) -0.446*** 1.477*** -0.081 0.035 (0.093) (0.268) (0.055) (0.046) -0.411*** 1.489*** -0.087* 0.036 (0.078) (0.224) (0.046) (0.038) -0.389*** 1.500*** -0.091** 0.038 (0.069) (0.199) (0.041) (0.034) -0.382*** 1.506*** -0.093** 0.038 (0.064) (0.185) (0.038) (0.031) Adj.R2 0.099 0.132 0.159 0.137 0.098 0.376 0.351 0.450 0.520 0.560 N 124 132 133 133 133 54 56 56 56 56

Adj PerAi denotes the industry-adjusted performance of the combined firm in the ith year following the acquisition. Adj PerA1Ai denotes the average industry-adjusted performance of the combined firm from the first to the ith year following the acquisition. Adj PerB1B3 denotes the average industryadjusted performance of the combined firm from the third to the first year prior to the acquisition. SizeA denotes the industry-adjusted size of the acquirer and is measured at the end of the first year prior to the acquisition. SizeT denotes the industry-adjusted size of the target and is measured at the end of the first year prior to the acquisition. ***, **, * statistical significance at 1%, 5% and 10%.

34

Table 13: Comparison of adjusted values of efficiency prior to the M&A deal
Panel A. Acquirer. Industry-adjusted values for cost efficiency and profit efficiency before the deal (adjustment: mean by year and country)
N. of years before the deal

N 694 661 586 510 435 378

Cost efficiency
Mean (Std. Dev) % +ve cases (Z-test) Mean (Std. Dev)

Profit efficiency
% +ve cases (Z-test)

B1 B2 B3 B4 B5 B6

.0145 (.1200) .0214 (.1218) .0285 (.1422) .0264 (.1355) .0203 (.1343) .0229 (.1341)

57%(-4.25) 59% (-5.16) 58% (-4.58) 58% (-4.51) 53% (-3.27) 56% (-3.16)

.0739 (.1062) .0722 (.1061) .0761 (.1117) .0794 (.1104) .0724 (.1053) .0822 (.0933)

83% (-17.38) 85% (-16.68) 85% (-16.03) 88% (-15.88) 85% (-13.96) 91% (-14.78)

Panel B. Target. Industry-adjusted values for cost efficiency and profit efficiency before the deal (adjustment: mean by year and country)
N. of years before the deal

N 264 251 211 181 160 130

Cost efficiency
Mean (Std. Dev) % +ve cases (Z-test) Mean (Std. Dev)

Profit efficiency
% +ve cases (Z-test)

B1 B2 B3 B4 B5 B6

.0026 (.1313) -.0014 (.1374) .0113 (.1458) .0119 (.1416) .0195 (.1565) .0212 (.1285)

56% (-1.949) 57% (-1.764) 59% (-2.737) 59% (-2.289) 67% (-3.169) 64% (-2.594)

.0535 (.1208) .0596 (.1160) .0658 (.1177) .0727 (.1338) .0708 (.1439) .0657 (.1327)

76% (-7.746) 78% (-8.220) 81% (-8190) 78% (-7.669) 79% (-6.854) 83% (-6.422)

Panel C. Acquirer versus Target. Industry-adjusted values for cost efficiency and profit efficiency before the deal (adjustment: mean by year and country)
N. of years before the deal

Cost efficiency
Mean (Std. Dev) % +ve cases (Z-test) Mean (Std. Dev)

Profit efficiency
% +ve cases (Z-test)

.0093 45% .0019 47% (.1625) (-.561) (.1443) (.647) .0179 45% .0038 44% B2 188 (.1710) (-1.052) (.1436) (-.652) .0088 43% .0049 42% B3 157 (.1798) (-.132) (.1538) (-.629) .0122 45% -.0019 45% B4 123 (.1807) (-.494) (.1667) (-.303) 42% .0036 42% B5 .0080 100 (.1839) (-.283) (.1562) (-.175) -.0025 48% .0133 41% B6 89 (.1717) (-.290) (.1452) (-.040) % of positive cases under the Wilcoxon test. , , Z-test respectively statistically significant at 1%, 5% and 10%. Total number of deals: 647. Number of domestic deals: 345. Number of cross-border deals: 302. B1

196

35

Table 14: Comparison of X-efficiency before and after the deal

Panel A. Post values of the combined bank vs. Pre values of the merging banks. Adjusted for cost and profit efficiency (adjustment: mean by year and country)
No of years after the deal Cost efficiency Cost efficiency Base year: B6B1 Base year: B3B1 Mean % negative cases Mean % negative cases (Std. Dev) (Z-test) (Std. Dev) (Z-test) Profit efficiency Base year: B6B1 Mean % positive cases (Std. Dev) (Z-test) Profit efficiency Base year: B3B1 Mean % positive cases (Std. Dev) (Z-test)

A1 n. deals: 160 A2 n. deals: 136 A3 n. deals: 121 A4 n. deals: 104 A5 n. deals: 77 A6 n. deals: 49

0.0308*** (0.082) 0.0366*** (0.074) 0.0375*** (0.079) 0.0397*** (0.083) 0.0407*** (0.093) 0.0510*** (0.104)

60% 0.0302*** (-4.177) (0.079) 71% 0.0358*** (-5.467) (0.073) 65% 0.0364*** (-5.053) (0.077) 73% 0.0387*** (-4.887) (0.080) 65% 0.0386*** (-3.465) (0.091) 80% 0.0436*** (-3.576) (0.106)

64% (-4.519) 72% (-5.629) 69% (-5.308) 74% (-5.034) 72% (-3.614)

-0.0117* (0.084) -0.0187** (0.087) -0.0155** (0.085) -0.0225** (0.088) -0.0424** (0.124)

53% (-1.603) 52% (-1.616) 52% (-1.380) 54% (-2.276) 65% (-3.064)

-0.0057 (0.081) -0.0127* (0.082) -0.0092* (0.079) -0.0155* (0.082) -0.0340** (0.121)

52% (-0.829) 46% (0.419) 44% (-0.317) 46% (-1.211) 61% (-2.459) 73% (-3.344)

75% -0.0533*** (-3.057) (0.094)

73% -0.0444*** (-3.566) (0.087)

Mean (A1A3) 0.0380 0.0734 -0.0073 0.0805 Mean (A1A6) 0.0400 0.0746 -0.0170 0.0861 Panel B. Domestic M&A. Post values of the combined bank vs. Pre values of the merging banks. Adjusted for cost and profit efficiency (adjustment: mean by year and country)
No of years After the deal Cost efficiency Cost efficiency Base year: B6B1 Base year: B3B1 Mean %negative cases Mean % negative cases (Std. Dev) (Z-test) (Std. Dev) (Z-test) Profit efficiency Base year: B6B1 Mean % positive cases (Std. Dev) (Z-test) Profit efficiency Base year: B3B1 Mean % positive cases (Std. Dev) (Z-test)

A1 n. deals: 87 A2 n. deals: 70 A3 n. deals: 64 A4 n. deals: 55 A5 n. deals: 41

0.0384*** (0.094) 0.0450*** (0.076) 0.0461*** (0.073) 0.0460*** (0.078) 0.0449*** (0.088)

57% 0.0387*** (-2.963) (0.090) 70% 0.0441*** (-4.322) (0.073) 70% 0.0444*** (-4.588) (0.070) 76% 0.0432*** (-4.223) (0.075) 76% 0.0397*** (-3.285) (0.085)

62% (-3.318) 70% (-4.427) 73% (-4.628) 75% (-4.198) 76% (-3.065)

-0.0055 (0.081) -0.0068 (0.077) -0.0017 (0.072) -0.0137 (0.075) -0.0397* (0.148)

54% (-0.601) 44% (-0.243) 44% (-0.187) 49% (-1.089) 59% (-1.432)

-0.0022 (0.081) -0.0048 (0.074) 0.0001 (0.068) -0.0106 (0.068) -0.0333 (0.145)

50% (-0.76) 37% (-0.688) 38% (-0.983) 42% (-0.369) 59% (-1.160)

A6 0.0544** 86% 0.0463** 82% -0.0521*** 68% -0.0452*** 68% n. deals: 28 (0.106) (-3.006) (0.104) (-2.983) (0.084) (-2.788) (0.075) (-2.801) Panel C. Cross-border M&A. Post values of the combined bank vs. Pre values of the merging banks. Adjusted for cost and profit efficiency (adjustment: mean by year and country)
Number of years After the deal Cost efficiency Cost efficiency Base year: B6B1 Base year: B3B1 Mean % negative cases Mean % negative cases (Std. Dev) (Z-test) (Std. Dev) (Z-test) Profit efficiency Base year: B6B1 Mean % positive cases (Std. Dev) (Z-test) Profit efficiency Base year: B3B1 Mean % positive cases (Std. Dev) (Z-test)

A1 n. deals: 73 A2 n. deals: 66 A3 n. deals: 57 A4 n. deals: 49 A5 n. deals: 36 A6 n. deals: 21

0.0217*** (0.065) 0.0276*** (0.072) 0.0278** (0.085) 0.0327** (0.089) 0.0359** (0.100) 0.0465* (0.104)

64% 0.0199*** (-2.966) (0.063) 71% 0.0269*** (-3.440) (0.072) 58% 0.0273** (-2.411) (0.083) 69% 0.0336*** (2.591) (0.087) 53% 0.0374** (-1.650) (0.099) 71% 0.0397 (-1.929) (0.110)

65% (-3.143)

-0.0190* (0.088)

52% (-1.685) 59% (-2.399) 61% (-2.054) 59% (-2.084) 72% (-2.765) 81% (-2.416)

-0.0099 (0.081) -0.0213* (0.089) -0.0199* (0.089) -0.0212 (0.095) -0.0349** (0.086) -0.0433* (0.103)

54% (-1.038) 55% (-1.650) 52% (-1.224) 50% (-1.272) 63% (-2.260) 80% (-2.016)

74% -0.0313*** (-3.683) (0.095) 64% -2.847) 73% (-2.903) -0.0311** (0.096) -0.0323** (0.101)

69% -0.0455*** (-2.047) (0.090) 65% -0.0549** (-1.456) (0.108)

Base year are weighted averages of the performance measure in the years prior to the M&A of the target and acquiring banks. ***, **, * T-test respectively statistically significant at 1%, 5% and 10%. , , Z-test respectively statistically significant at 1%, 5% and 10%. Total number of deals: 647. Number of domestic deals: 345. Number of cross-border deals: 302.

Table 15: X-efficiency before and after the deal: M&A impact and trend
Panel A: M&A sample (post vs. 3 years pre- deal) Panel B: M&A sample (post vs. 6 years pre- deal) R R2 AdjR2

Post, Pre
A1, B3B1 A2, B3B1 A3, B3B1 A4, B3B1 A5, B3B1 A6, B3B1 A1A3, B3B1 A1, B3B1 A2, B3B1 A3, B3B1 A4, B3B1 A5, B3B1 A6, B3B1 A1A3, B3B1

Post, Pre
A1, B6B1 A2, B6B1 A3, B6B1 A4, B6B1 A5, B6B1 A6, B6B1 A1A6, B6B1 A1, B6B1 A2, B6B1 A3, B6B1 A4, B6B1 A5, B6B1 A6, B6B1 A1A6, B6B1

R2

AdjR2

Cost efficiency 0.029*** (0.006) 0.033*** (0.006) 0.031*** (0.006) 0.031*** (0.007) 0.027*** (0.008) 0.020* (0.012) 0.036*** (0.006) 0.001 (0..8) -0.003 (0.009) 0.005 (0.009) -0.012 (0.011) -0.012 (0.019) -0.024 (0.030) 0.007 (0.007) 0.214** (0.087) 0.297*** (0.092) 0.85*** (0.092) 0.299*** (0.108) 0.529*** (0.180) 0.621*** (0.173) 0.796*** (0.062) 0.689*** (0.057) 0.558*** (0.063) 0.554*** (0.071) 0.462*** (0.075) 0.349*** (0.099) 0.777*** (0.057) 0.715 0.726 0.630 0.616 0.583 0.463 0.735 0.512 0.527 0.397 0.389 0.340 0.214 0.541

AdjPer M & A , post = + AdjPer M & A , pre +

0.509 0.523 0.392 0.373 0.331 0.197 0.538

0.029*** (0.006) 0.033*** (0.006) 0.032*** (0.006) 0.031*** (0.007) 0.027*** (0.008) 0.023** (0.012) 0.038*** (0.006) -0.001 (0.008) -0.004 (0.010) 0.007 (0.010) -0.008 (0.012) -0.010 (0.020) -0.014 (0.020) 0.004 (0.008) 0.270*** (0.090) 0.352*** (0.097) 0.330*** (0.097) 0.347*** (0.115) 0.575*** (0.183) 0.583*** (0.184)

0.761*** (0.063) 0.670*** (0.056) 0.527*** (0.063) 0.528*** (0.071) 0.440*** (0.075) 0.354*** (0.097) 0.746*** (0.056) 0.855*** (0.069) 0.813*** (0.077) 0.725*** (0.082) 0.826*** (0.098) 0.662*** (0.151) 0.571*** (0.164) 0.693*** (0.067)

0.691 0.715 0.608 0.595 0.561 0.471 0.724

0.478 0.512 0.370 0.354 0.314 0.222 0.524

0.474 0.508 0.365 0.347 0.305 0.205 0.521

Profit efficiency 0.887*** 0.729 0.531 (0.067) 0.874*** 0.716 0.512 (0.074) 0.818*** 0.674 0.454 (0.083) 0.952*** 0.696 0.484 (0.952) 0.743*** 0.475 0.226 (0.160) 0.751*** 0.504 0.254 (0.190) 0.774*** 0.691 0.477 (0.064) Profit efficiency and size 0.907*** (0.066) 0.900*** (0.072) 0.866*** (0.080) 1.012*** (0.096) 0.903*** (0.160) 1.171*** (0.207) -0.008* (0.005) -0.017*** (0.005) -0.018*** (0.005) -1.018*** (0.006) -0.031*** (0.010) -0.032*** (0.009)

AdjPerM & A , post = + AdjPerM & A, pre +

0.528 0.508 0.449 0.479 0.215 0.238 0.474

0.701 0.673 0.628 0.641 0.450 0.453 0.633

0.492 0.453 0.395 0.411 0.203 0.206 0.401

0.488 0.448 0.390 0.406 0.192 0.189 0.397

AdjPer M & A , post = + AdjPer M & A , pre + Size A, pre + Size T , pre +

A1, B3B1 A2, B3B1 A3, B3B1 A4, B3B1 A5, B3B1 A6, B3B1

-0.005 (0.004) 0.000 (0.004) 0.003 (0.004) 0.000 (0.004) 0.001 (0.007) -0.007 (0.006)

AdjR2 0.540 0.543 0.496 0.515 0.294 0.399

A1, B6B1 A2, B6B1 A3, B6B1 A4, B6B1 A5, B6B1 A6, B6B1

0.892*** (0.069) 0.855*** (0.075) 0.784*** (0.080) 0.897*** (0.097) 0.834*** (0.153) 0.906*** (0.188)

-0.009* (0.005) -0.018*** (0.005) -0.019*** (0.005) -0.018*** (0.007) -0.031*** (0.010) -0.027*** (0.009)

-0.006* (0.004) -0.002 (0.004) 0.001 (0.004) -0.003 (0.005) -0.003 (0.007) -0.008 (0.007)

AdjR2 0.512 0.495 0.443 0.448 0.277 0.315

Panel C: Domestic M&A sample (post vs. 6 years pre- deal) 0.033*** (0.010) 0.032*** (0.009) 0.034*** (0.008) 0.025*** (0.009) 0.011 (0.012) 0.005 (0.018) 0.026*** (0.009) 0.036*** (0.011) 0.035*** (0.011) 0.024** (0.012) -0.007 (0.032) -0.009 (0.021) 0.780*** (0.107) 0.600*** (0.089) 0.542*** (0.090) 0.458*** (0.099) 0.359*** (0.114) 0.267* (0.150) 0.427*** (0.097) 0.358*** (0.110) 0.436*** (0.113) 0.454*** (0.121) 0.544** (0.304) 0.485 (0.184 R 0.622 0.633 0.607 0.538 0.450 0.329 0.431 0.367 0.441 0.459 0.275 0.459 R2 0.386 0.400 0.368 0.289 0.203 0.109 0.186 0.135 0.194 0.211 0.076 0.211

A1, B3B1 A2, B3B1 A3, B3B1 A4, B3B1 A5, B3B1 A6, B3B1 A1, B3B1 A2, B3B1 A3, B3B1 A4, B3B1 A5, B3B1 A6, B3B1

Panel D: Cross-border M&A sample (post vs. 6 years pre- deal) Cost efficiency AdjR2 R R2 0.379 A1, B6B1 0.024*** 0.757*** 0.791 0.626 (0.007) (0.070) 0.392 A2, B6B1 0.032*** 0.727*** 0.763 0.582 (0.008) (0.077) 0.358 A3, B6B1 0.030*** 0.521*** 0.604 0.365 (0.009) (0.093) 0.276 A4, B6B1 0.034*** 0.566*** 0.610 0.372 (0.011) (0.107) 0.182 A5, B6B1 0.042*** 0.431*** 0.563 0.317 (0.013) (0.108) 0.074 A6, B6B1 0.039** 0.369** 0.545 0.297 (0.016) (0.130) Profit efficiency 0.176 A1, B6B1 -0.029** 1.107*** 0.819 0.670 (0.013) (0.092) 0.122 A2, B6B1 -0.032** 1.008*** 0.774 0.600 (0.015) (0.103) 0.181 A3, B6B1 -0.020 0.891*** 0.709 0.503 (0.018) (0.11) 0.196 A4, B6B1 -0.040** 1.071*** 0.729 0.531 (0.021) (0.147) 0.052 A5, B6B1 -0.011 0.712*** 0.648 0.420 (0.022) (0.143) 0.180 A6, B6B1 -0.021 0.676** 0.450 0.203 (0.039) (0.307)

AdjR2 0.620 0.576 0.354 0.359 0.297 0.259 0.666 0.594 0.494 0.521 0.403 0.161

Post values of the combined bank vs. Pre values of the merging banks. Adjusted values for cost and profit efficiency (adjustment: mean by year and country).

37

Table 16: Descriptive statistics of institutional, deal- specific and bank-specific determinants of the change in X-efficiency
N. obs. Minimum Maximum Mean Std. Deviation T_Freedom from Government 312 .00 0.90 .3901 .1841 T_Regulatory quality 291 -.65 1.94 .9469 .4243 T_Concentration index 367 .00 0.80 .2607 .1987 A_Freedom from Government 703 .01 0.94 .3927 .1388 A_ Regulatory quality 631 .02 1.94 1.0737 .3191 A_Concentration index 558 .00 0.80 .2553 .2064 Payment method (=1 if Cash only) 970 .00 1.00 .5515 .4976 Deal Period: 2000-2005 970 .00 1.00 .4701 .4994 Deal Period: 1994-1999 970 .00 1.00 .5299 .4994 Deal Period: 1991-1993 970 .00 1.00 .1309 .3375 Cross border dummy (=1 if cross border) 970 .00 1.00 .8557 .3516 C_big 708 .00 1.00 .3319 .4712 C_medium 708 .00 1.00 .3362 .4727 A_big 786 .00 1.00 .3282 .4699 A_medium 786 .00 1.00 .3384 .4735 T_medium 303 .00 1.00 .3333 .4722 T_small 271 .00 1.00 .2435 .4300 C_ Traditional banking 708 .06 .90 .5269 .1322 A_Traditional banking 786 .02 .89 .5132 .1252 T_ Traditional banking 303 .06 .96 .5461 .1743 C_ Risk 636 .00 3.41 .0591 .1729 A_Risk 717 .00 3.41 .0569 .1965 T_ Risk 266 .00 1.05 .0899 .1444 Freedom from government (http://www.heritage.org/research/features/index/) is defined to include all government expenditures - including consumption and transfers - and state-owned enterprises. Regulatory quality (www.worldbank.org), the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development. Concentration index = Total value of assets of the five biggest banks (CR5). Traditional banking = Loans/Total assets; Risk = Stand. Dev ROE; Big = Fist tertile (ln (Total assets)); Medium = Second tertile (ln (Total assets)); Small = Third tertile (ln (Total assets)).

Table 17: Determinants of changes in X-efficiency prior and after M&A

AdjPerM & A, pre vs post = + (CV A, pre , CVT , pre , CVC , post )
Independent Variables () and intercept (): Intercept T_Freedom from Government T_Regulatory quality T_Concentration index A_Freedom from Government A_ Regulatory quality A_Concentration index Payment method dummy (=1 if Cash only) Deal Period: 2000-2005 Deal Period: 1994-1999 Cross border dummy (=1 if cross border) C_big C_medium A_big A_medium T_medium T_small C_Traditional banking A_Traditional banking T_ Traditional banking C_Risk A_Risk T_Risk N. of obs. R2 Change in Profit efficiency Par T-stat 0.141** 0.064 0.204** 0.057* 0.023 -0.337*** -0.072** -0.142*** -0.024* -0.12 -0.049*** 0.023 0.005 -0.020 -0.076** -0.030 -0.034* -0.002 0.352*** -0.408*** 0.040 1.096*** 0.431*** 0.230** 96 0.793 0.096 0.035 0.049 0.087 0.034 0.054 0.014 0.023 0.016 0.017 0.036 0.024 0.030 0.024 0.018 0.020 0.101 0.101 0.055 0.248 0.129 0.104 Change in Cost efficiency Par T-stat 0.159*** 0.045 0.365*** 0.105 -0.030 -0.317*** 0.065** -0.009 -0.022* -0.083*** -0.026** 0.006 -0.004 0.010 -0.002 -0.010 -0.019 -0.005 0.000 -0.002** 0.001 0.661** 0.314*** 0.017 96 0.665 0.079 0.029 0.044 0.073 0.028 0.045 0.012 0.020 0.011 0.692 0.905 0.020 0.026 0.020 0.015 0.017 0.001 0.001 0.001 0.269 0.105 0.094

38

Table 18: Characteristics of successful vs. unsuccessful deals

Profit efficiency Characteristics Size Capital Risk Diversification Freedom from government Regulatory quality Successful 17.810 (1.567) 0.209 (0.177) 0.055 (0.079) 1.218 (2.013) 0.397 (0.081) 1.022 (0.276) Acquirer Unsuccessful 18.779 (0.863) 0.279 (0.118) 0.024 (0.023) 0.634 (0.612) 0.445 (0.116) 1.190 (0.293) Acquirer Unsuccessful Difference in means -0.969*** [-2.968] -0.070* [-1.812] 0.031** [1.965] 0.584 [1.517] -0.048* [-1.660] -0.168** [-1.995] Cost efficiency Successful 16.142 (1.726) 0.156 (0.213) 0.1729 (0.201) 0.276 (0.517) 0.441 (0.243) 0.994 (0.407) Target Unsuccessful 16.399 (1.995) 0.193 (0.252) 0.070 (0.079) 0.549 (0.599) 0.437 (0.224) 0.713 (0.387) Difference in means -0.257 [-0.533] -0.036 [-0.606] 0.103** [2.449] -0.273* [-1.640] 0.005 [0.066] 0.280** [2.391]

Target Difference in Successful Unsuccessful Difference in means means Size 18.143 17.309 0.833** 16.751 16.194 0.557 (1.272) (1.729) [2.127] (2.272) (1.947) [1.020] Capital 0.197 0.136 0.060* 0.273 0.180 0.093 (0.148) (0.117) [1.746] (0.302) (0.240) [1.327] Loan loss 0.295 0.165 0.131** 0.242 0.128 0.114* provision (0.310) (0.160) [2.053] (0.308) (0.170) [1.776] Deal period: 0.533 0.833 -0.300*** 0.533 0.833 -0.300*** 1994-1999 (0.507) (0.379) [-2.594] (0.507) (0.379) [-2.594] Size = ln (Total assets). Capital = (Equityi Equitymin)/Equitymax. Diversification = Off-balance sheet /total assets. Loan loss provision (llp) = (llpi llpmin)/llpmax. Standard deviations in ( ). T-values in [ ]. ***, **, * T-test respectively statistically significant at 1%, 5% and 10%. Characteristics Successful

39